DETAILED ACTION
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-8, 15-23, and 31-33 are rejected under 35 U.S.C. 103 as being unpatentable over Gibson (US 20240020950 A1), and in view of Veronesi et al. (US 20230102406 A1), and further in view of Jha (US 20210209424 A1).
Re Claim 1: Gibson discloses a method of training a machine learning model to identify an expected distribution of mapping categories for a geographic area (see Gibson: e.g., -- computer-implemented methods and systems are provided for pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system. To do this, the amount that each class is contained within each area image relative to the whole dataset is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken-, in abstract, and, -- computer-implemented methods and systems for generating training data for training a machine learning system to automatically extract topographic features. In particular, aspects described herein relate to a method of pre-processing image data to generate improved training data.--, in [0001], and, -- pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system, with each image containing a part of the geographic region of interest. To do this, the amount that each class is contained within each area image relative to the whole dataset (i.e., the whole geographic region captured) is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken. This ensures that the distribution of patches taken from each area prioritises features that are less common….. a computer-implemented method of processing image data for use in training a machine learning system for classifying image data, the method comprising obtaining image data comprising a plurality of images, each image corresponding to a respective geographic area, processing the image data to identify one or more classes of topographic feature contained within each image, wherein processing the image data comprises determining a quantity of each respective topographic feature class contained within each image, generating a first dataset comprising a set of values for each image, wherein each value is representative of the quantity of one of the plurality of topographic feature classes contained within the respective image, and processing the first dataset to determine a number of samples required from each image for each topographic feature class. Processing the first dataset comprises (i) generating a normalised dataset, wherein the values of the first dataset are normalised across each image and each topographic feature class, and (ii) for each topographic feature class, calculating the number of samples required from each image based on the normalised dataset and a target number of samples for the respective topographic feature class. --, in [0008]-[0010]);
Gibson however does not explicitly disclose to identify an expected distribution of mapping categories for a geographic area at a given time,
Veronesi discloses that training a machine learning model to identify an expected distribution of mapping categories for a geographic area at a given time (see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]);
Veronesi and Gibson are combinable as they are in the same field of endeavor: machine learning algorithms used in geographic mapping and images processing. Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to further modify Gibson’s method using Veronesi’s teachings by including training a machine learning model to identify an expected distribution of mapping categories for a geographic area at a given time to Gibson’ geographic mapping in order to apply a machine learning model using the feature vector to generate geographic mapping for each of the plurality of pixels of the images for the right time and high accuracy (see Veronesi: e.g., in abstract, [00003]-[0005], [0021], [0025]-[0028], and [0058]);
Gibson as modified by Veronesi further disclose receiving, by processing circuitry, a plurality of maps of the geographic area, wherein: each map of the plurality of maps is generated based on a respective plurality of overhead images (see Gibson: e.g., -- [0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. {classes read on categories} A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.--, in [0002]-[0005], and [0035]-[0037]),
Gibson as modified by Veronesi however still do not explicitly disclose that a plurality of overhead images captured during a respective portion of a time period,
Jha discloses a plurality of overhead images captured during a respective portion of a time period (see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]);
Gibson (as modified by Veronesi) and Jha are combinable as they are in the same field of endeavor: machine learning algorithms used in geographic mapping and images processing. Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to further modify Gibson (as modified by Veronesi)’s method using Jha’s teachings by including a plurality of overhead images captured during a respective portion of a time period to Gibson (as modified by Veronesi)’s geographic mapping in order to determine a quantitative relative change in classes of the geographic area over a time duration from the first time frame to the second time frame (see Jha: e.g., in abstract, [00006]-[0009], and [0016]-[0019]);
Gibson as modified by Veronesi and Jha further disclose each overhead image of the respective pluralities of overhead images comprises a respective plurality of pixels, and each pixel is designated as being of a particular mapping category of a plurality of mapping categories (see Gibson: e.g., -- [0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. {classes read on categories} A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.--, in [0002]-[0005], and [0035]-[0037];
also see Veronesi: e.g., --[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]);
training, using the processing circuitry and the plurality of maps, the machine learning model to identify the expected distribution for the mapping categories of the geographic area at given time (see Gibson: e.g., -- [0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. {classes read on categories} A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.--, in [0002]-[0005], and [0035]-[0037];
also see Veronesi: e.g., --[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]).
Re Claim 2, Gibson as modified by Veronesi and Jha further disclose receiving, by the processing circuitry and for each geographic area of a plurality of geographic areas, a respective plurality of maps, wherein:
for each geographic area of the plurality of geographic areas, the respective plurality of maps is generated based on a respective plurality of overhead images captured during a respective portion of a time period (see Gibson: e.g., -- [0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. {classes read on categories} A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.--, in [0002]-[0005], and [0035]-[0037];
also see Veronesi: e.g., --[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028];
further see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]);
each pixel of each respective plurality of overhead images is designated as being of a particular mapping category of a plurality of mapping categories (see Gibson: e.g., -- [0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. {classes read on categories} A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.--, in [0002]-[0005], and [0035]-[0037];
also see Veronesi: e.g., --[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]);
wherein training the machine learning model is further performed using the respective pluralities of maps for the plurality of geographic areas, to identify an expected distribution for mapping categories of a particular geographic area of the plurality of geographic areas at a particular time (see Gibson: e.g., -- [0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. {classes read on categories} A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.--, in [0002]-[0005], and [0035]-[0037];
also see Veronesi: e.g., --[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]).
Re Claim 3, Gibson as modified by Veronesi and Jha further disclose wherein training the machine learning model causes the machine learning model to learn a temporal distribution of mapping categories and confidences observed for the geographic area over the time period see Veronesi: e.g., --[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028];
further see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]); and
the method further comprises generating the model, a vector, based on the temporal distribution of mapping categories and confidences observed for geographic area over the time period, summarizing variability of mapping categories of the geographic area over the time period (see Gibson: e.g., -- [0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. {classes read on categories} A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.--, in [0002]-[0005], and [0035]-[0037]).
Re Claim 4, Gibson as modified by Veronesi and Jha further disclose wherein the time period corresponds to: a particular year, and each respective portion of the time period corresponds to a month of the particular year (see Veronesi: e.g., --[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028];
further see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]);
a plurality of seasons comprising winter, spring, summer, and autumn, and each respective portion of the time period corresponds to a respective season of the plurality of seasons; or a plurality of years, and each respective portion of the time period corresponds to a respective year of the plurality of years see Veronesi: e.g., --[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028];
further see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]).
Re Claim 5, Gibson as modified by Veronesi and Jha further disclose wherein the machine learning model is a statistical machine learning model, a Bayesian probabilistic model, a random forest model, a deep learning model, a transformer model using attention heads, or a combination thereof (see Gibson: e.g., --[0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.
[0038] Aspects described herein seek to improve the generation of training data by providing methods of optimizing the distribution of sample areas, wherein the patches are selected by both region and class, and methods of optimizing the patch selection to minimize class imbalance within each patch.
Distribution of Patches [0039] A problem with geographic distribution of training data is that examples of some classes only appear in certain areas. For example, railway lines will not appear in a large number of images, compared to buildings or vegetation. A random distribution actively biases against these classes and would result in a small number of patches being used to determine an entire class.
[0040] Additionally, different areas have different signatures (for example, buildings in the business district of a city will look different to those in a residential area), and so it is important to select patches from different locations across a wide variety of different classes, as patches from the same area tend to train the machine learning system to only detect features from that location.
[0041] The distribution method described herein thus ensures that the patches fed into the model have at least some of each classes in equal proportion, and ensures that the existence of the class is balanced across the set of patches. This helps to smooth and stabilize the minority classes and ensures the model actively trains on the class across thousands of examples rather than a few hundred.--, in [0035]-[0041], and
--[0045] For example, for a dataset of 10000 patches and 5 classes, it may be necessary that each class exists in at least 2000 of the patches. This means that the model is exposed to the class during gradient descent even if only in a few pixels. Any shortfalls due to saturation can be made up, for example, using data augmentation techniques, which will be targeted at the relevant class rather than being random. Likewise, if that dataset of 10000 patches corresponds to 100 regions, approximately 100 patches will have been taken from each region.
[0046] In known methods, random sampling is used to tile up an image into patches, and this often biases against certain classes, leaving only a hundred patches to train from out of a dataset that may comprise 10000 patches.
[0047] Therefore, an objective is to determine a distribution within the training dataset that best represents the geographic diversity but also enforces a minimum frequency with which classes will appear within the patch dataset.--, in [0045]-[0047];
also see Veronesi: e.g., --[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028];
further see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]);
.
Re Claim 6, Gibson as modified by Veronesi and Jha further disclose using auxiliary geospatial data as a parameter during training of the machine learning wherein the auxiliary geospatial data comprises at least one of weather data, topographic data, or signal emissions data for the geographic area (see Gibson: e.g., -- computer-implemented methods and systems are provided for pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system. To do this, the amount that each class is contained within each area image relative to the whole dataset is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken-, in abstract, and, -- computer-implemented methods and systems for generating training data for training a machine learning system to automatically extract topographic features. In particular, aspects described herein relate to a method of pre-processing image data to generate improved training data.--, in [0001], and, -- pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system, with each image containing a part of the geographic region of interest. To do this, the amount that each class is contained within each area image relative to the whole dataset (i.e., the whole geographic region captured) is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken. This ensures that the distribution of patches taken from each area prioritises features that are less common….. a computer-implemented method of processing image data for use in training a machine learning system for classifying image data, the method comprising obtaining image data comprising a plurality of images, each image corresponding to a respective geographic area, processing the image data to identify one or more classes of topographic feature contained within each image, wherein processing the image data comprises determining a quantity of each respective topographic feature class contained within each image, generating a first dataset comprising a set of values for each image, wherein each value is representative of the quantity of one of the plurality of topographic feature classes contained within the respective image, and processing the first dataset to determine a number of samples required from each image for each topographic feature class. Processing the first dataset comprises (i) generating a normalised dataset, wherein the values of the first dataset are normalised across each image and each topographic feature class, and (ii) for each topographic feature class, calculating the number of samples required from each image based on the normalised dataset and a target number of samples for the respective topographic feature class. --, in [0008]-[0010];
also see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]).
Re Claim 7, Gibson as modified by Veronesi and Jha further disclose determining one or more trends for the mapping categories of the geographic area over the time period (see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]),
wherein training the machine learning model is further based on the determined one or more trends (see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]).
Re Claim 8, Gibson as modified by Veronesi and Jha further disclose determining a set of hyperparameters of the machine learning model based on identifying spatial segments and temporal segments summarizing the respective pluralities of overhead image ((see Gibson: e.g., -- [0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. {classes read on categories} A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.--, in [0002]-[0005], and [0035]-[0037]; also see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]).
Re Claim 15, Gibson discloses a method comprising: identifying a distribution of mapping categories for a geographic area, based on at least one overhead image of the geographic area (see Gibson: e.g., -- computer-implemented methods and systems are provided for pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system. To do this, the amount that each class is contained within each area image relative to the whole dataset is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken-, in abstract, and, -- computer-implemented methods and systems for generating training data for training a machine learning system to automatically extract topographic features. In particular, aspects described herein relate to a method of pre-processing image data to generate improved training data.--, in [0001], and, -- pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system, with each image containing a part of the geographic region of interest. To do this, the amount that each class is contained within each area image relative to the whole dataset (i.e., the whole geographic region captured) is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken. This ensures that the distribution of patches taken from each area prioritises features that are less common….. a computer-implemented method of processing image data for use in training a machine learning system for classifying image data, the method comprising obtaining image data comprising a plurality of images, each image corresponding to a respective geographic area, processing the image data to identify one or more classes of topographic feature contained within each image, wherein processing the image data comprises determining a quantity of each respective topographic feature class contained within each image, generating a first dataset comprising a set of values for each image, wherein each value is representative of the quantity of one of the plurality of topographic feature classes contained within the respective image, and processing the first dataset to determine a number of samples required from each image for each topographic feature class. Processing the first dataset comprises (i) generating a normalised dataset, wherein the values of the first dataset are normalised across each image and each topographic feature class, and (ii) for each topographic feature class, calculating the number of samples required from each image based on the normalised dataset and a target number of samples for the respective topographic feature class. --, in [0008]-[0010]);
Gibson however does not explicitly disclose overhead image of the geographic area captured at a given time,
Veronesi discloses overhead image of the geographic area captured at a given time (see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]);
Veronesi and Gibson are combinable as they are in the same field of endeavor: machine learning algorithms used in geographic mapping and images processing. Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to further modify Gibson’s method using Veronesi’s teachings by including overhead image of the geographic area captured at a given time to Gibson’ geographic mapping in order to apply a machine learning model using the feature vector to generate geographic mapping for each of the plurality of pixels of the images for the right time and high accuracy (see Veronesi: e.g., in abstract, [00003]-[0005], [0021], [0025]-[0028], and [0058]);
Gibson as modified by Veronesi and Jha further disclose inputting indications of the given time and the geographic area to a trained machine model (see Gibson: e.g., -- [0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. {classes read on categories} A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.--, in [0002]-[0005], and [0035]-[0037]; also see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]);
outputting, using the trained machine learning model, an expected distribution of the mapping categories of the geographic area at the given time (see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]; further see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]);
comparing the identified distribution of mapping categories for the geographic area to the expected distribution of the mapping categories of the geographic area at the given time to identify difference between the identified distribution and the expected distribution (see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]; further see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]); and
determining whether the difference between the identified distribution and the expected distribution is an anomaly (see Gibson: e.g., -- computer-implemented methods and systems are provided for pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system. To do this, the amount that each class is contained within each area image relative to the whole dataset is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken-, in abstract, and, -- computer-implemented methods and systems for generating training data for training a machine learning system to automatically extract topographic features. In particular, aspects described herein relate to a method of pre-processing image data to generate improved training data.--, in [0001], and, -- pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system, with each image containing a part of the geographic region of interest. To do this, the amount that each class is contained within each area image relative to the whole dataset (i.e., the whole geographic region captured) is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken. This ensures that the distribution of patches taken from each area prioritises features that are less common….. a computer-implemented method of processing image data for use in training a machine learning system for classifying image data, the method comprising obtaining image data comprising a plurality of images, each image corresponding to a respective geographic area, processing the image data to identify one or more classes of topographic feature contained within each image, wherein processing the image data comprises determining a quantity of each respective topographic feature class contained within each image, generating a first dataset comprising a set of values for each image, wherein each value is representative of the quantity of one of the plurality of topographic feature classes contained within the respective image, and processing the first dataset to determine a number of samples required from each image for each topographic feature class. Processing the first dataset comprises (i) generating a normalised dataset, wherein the values of the first dataset are normalised across each image and each topographic feature class, and (ii) for each topographic feature class, calculating the number of samples required from each image based on the normalised dataset and a target number of samples for the respective topographic feature class. --, in [0008]-[0010]).
Re Claim 16, Gibson as modified by Veronesi and Jha further disclose wherein the trained machine learning model is trained using a plurality of maps of the geographic area, and wherein: each map of the plurality of maps is generated based on a respective plurality of overhead images captured during a respective portion of a time period (see Gibson: e.g., -- [0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. {classes read on categories} A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.--, in [0002]-[0005], and [0035]-[0037];
also see Veronesi: e.g., --[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]);
each overhead image of the respective pluralities of overhead images comprises a respective plurality of pixels, and each pixel is designated as being of a particular mapping category of a plurality of mapping categories (see Gibson: e.g., -- [0035] The machine learning system to be trained comprises a deep learning algorithm designed to semantically segment aerial or satellite images into land cover maps, which can then be processed further to output topographic maps. In this respect, the problem of creating maps from imagery treated as one of “semantic segmentation”, whereby each pixel within the image is allocated to a single class, from the set of classes that will be represented on the output land cover map, for example, building, road, water, railway, vegetation and the like. {classes read on categories} A post-processing step then generates the topographical map by “cleaning up” this output, for example, by polygonising continuous regions of pixels allocated to the same class and then regularizing those outlines into realistic shapes depending on the feature they represent, for example, so that buildings have straight edges.
[0036] It will be appreciated that any suitable machine learning model capable of classifying images may be used. For example, a convolutional neural network based on U-Net may be trained to perform semantic segmentation of aerial images using a gradient descent algorithm.
[0037] In order to train the machine learning system to accurately classify each pixel of a target image, it is important to use training data that has a balanced distribution of pixels for each of the classes that are to be used. Typically, as illustrated by FIG. 1, this is performed by capturing one or more test images 1 of different regions, and then selecting sample areas 10 (referred to herein as “patches”) within each test image 1 to label. In this respect, most test images 1 are too large to label entirely. To overcome this, each test image 1 will be divided into a grid of patches 10, and then patches 10 that represent different geographies, biomes or urban types are selected and labelled in order to obtain a set of patches, wherein each of the classes appear within at least one patch. Deciding how to select patches 10 so that all regions are represented well without giving undue preference to a given class can be very difficult to achieve and requires laborious manual inspection that is both expensive and time consuming.--, in [0002]-[0005], and [0035]-[0037];
also see Veronesi: e.g., --[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]).
Re Claim 17, Gibson as modified by Veronesi and Jha further disclose the comparing comprises identifying a plurality of candidate differences between the identified distribution and the expected distribution (see Gibson: e.g., -- computer-implemented methods and systems are provided for pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system. To do this, the amount that each class is contained within each area image relative to the whole dataset is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken-, in abstract, and, -- computer-implemented methods and systems for generating training data for training a machine learning system to automatically extract topographic features. In particular, aspects described herein relate to a method of pre-processing image data to generate improved training data.--, in [0001], and, -- pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system, with each image containing a part of the geographic region of interest. To do this, the amount that each class is contained within each area image relative to the whole dataset (i.e., the whole geographic region captured) is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken. This ensures that the distribution of patches taken from each area prioritises features that are less common….. a computer-implemented method of processing image data for use in training a machine learning system for classifying image data, the method comprising obtaining image data comprising a plurality of images, each image corresponding to a respective geographic area, processing the image data to identify one or more classes of topographic feature contained within each image, wherein processing the image data comprises determining a quantity of each respective topographic feature class contained within each image, generating a first dataset comprising a set of values for each image, wherein each value is representative of the quantity of one of the plurality of topographic feature classes contained within the respective image, and processing the first dataset to determine a number of samples required from each image for each topographic feature class. Processing the first dataset comprises (i) generating a normalised dataset, wherein the values of the first dataset are normalised across each image and each topographic feature class, and (ii) for each topographic feature class, calculating the number of samples required from each image based on the normalised dataset and a target number of samples for the respective topographic feature class. --, in [0008]-[0010]); and
determining whether the difference between the identified distribution and the expected distribution is an anomaly comprises ranking the plurality of candidate differences based on a degree to which a respective candidate difference of the plurality of candidate differences differs from the expected distribution (see Gibson: e.g., -- computer-implemented methods and systems are provided for pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system. To do this, the amount that each class is contained within each area image relative to the whole dataset is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken-, in abstract, and, -- computer-implemented methods and systems for generating training data for training a machine learning system to automatically extract topographic features. In particular, aspects described herein relate to a method of pre-processing image data to generate improved training data.--, in [0001], and, -- pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system, with each image containing a part of the geographic region of interest. To do this, the amount that each class is contained within each area image relative to the whole dataset (i.e., the whole geographic region captured) is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken. This ensures that the distribution of patches taken from each area prioritises features that are less common….. a computer-implemented method of processing image data for use in training a machine learning system for classifying image data, the method comprising obtaining image data comprising a plurality of images, each image corresponding to a respective geographic area, processing the image data to identify one or more classes of topographic feature contained within each image, wherein processing the image data comprises determining a quantity of each respective topographic feature class contained within each image, generating a first dataset comprising a set of values for each image, wherein each value is representative of the quantity of one of the plurality of topographic feature classes contained within the respective image, and processing the first dataset to determine a number of samples required from each image for each topographic feature class. Processing the first dataset comprises (i) generating a normalised dataset, wherein the values of the first dataset are normalised across each image and each topographic feature class, and (ii) for each topographic feature class, calculating the number of samples required from each image based on the normalised dataset and a target number of samples for the respective topographic feature class. --, in [0008]-[0010]).
Re Claim 18, Gibson as modified by Veronesi and Jha further disclose wherein the ranking further comprises: categorizing each respective candidate difference of the plurality of candidate differences based on a likelihood that the respective candidate difference is an anomaly, each categorization including an indication of an observed feature related to whether the respective candidate difference is an anomaly (see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]; further see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]).
Re Claim 19, Gibson as modified by Veronesi and Jha further disclose storing the plurality of candidate differences, and each respective categorization, in an unstructured database (see Veronesi: e.g., --[0027] Among the additional geospatial data, the system extracts elevation, slope, and aspect from databases such as the United States Geological Service (USGS) National Elevation Dataset and climate data (precipitation, temperature, and solar radiation) from ClimNA, which may be specific to North America. Soil data may also be included in the modeling from databases such as the gNATSGO database. Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.--, in [0027], and, -- [0033] Memory 214 may include a data collector 216, a data pre-processor 218, a feature vector generator 220, a machine learning model 222, a model trainer 224, a data post-processor 226, an overlay generator 228, and a normalization database 230. In brief overview, components 216-230 may cooperate to collect different types of data and images of a geographical region. Components 216-230 may generate a feature vector from data and the images and input the feature vector into a machine learning model that has been trained to output timber data for individual pixels of images. The machine learning model may output timber data for the image and components 216-230 may generate an interactive overlay from the timber data for display on a graphical user interface (GUI) 232.--, in [0031]-[0033], and, -- Data collector 216 may store the sets of measurements in a database (not shown) within forest inventory manager 206 to be used as labels in training datasets.--, in [0043]-[0045]).
Re Claim 20, Gibson as modified by Veronesi and Jha further disclose wherein identifying the difference between the identified distribution and the expected distribution further comprises: applying spatial clustering to segment spatially adjacent pixels of the at least one overhead image, likely to correspond to the difference between the identified distribution and the expected distribution, into an entity described by a vector (see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]);
Re Claim 21, Gibson as modified by Veronesi and Jha further disclose based on whether the difference is determined to be an anomaly, recommending performance of, or automatically performing, a follow-up observation of the geographic area (see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]; further see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]).
Re Claim 22, Gibson as modified by Veronesi and Jha further disclose wherein the recommending of the performance of, or the automatic performance of, the follow-up observation of the geographic area is performed based on receiving user feedback accepting the difference as an anomaly (see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]; further see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]).
Re Claim 23, Gibson as modified by Veronesi and Jha further disclose a type of the difference comprises a first mapping category indicated in the identified distribution for a portion of the geographic area being different from a second mapping category indicated in the expected distribution for the portion of the geographic area; and the follow-up observation is identified based on the type of the difference (see Veronesi: e.g., --the data processing system may train machine learning models to predict timber data for images over time and select the models that make the most accurate predictions to use in practice. For example, after inputting a series of training data sets into the machine learning models for training, the data processing system may evaluate the accuracy of the models by comparing the models' outputs against the expected values. The data processing system may select the machine learning model with the highest accuracy to use upon receiving a request to generate timber data for a geographical area.--, in [0058]; also see: --Filling intelligence gaps in-between surveys is a challenge as is getting ground resources to the right place at the right time to maximize impact. … providing an improved forest inventory by capturing a distribution and intermixing of different tree species within a forest and estimating a total volume and biomass of available timber in forest areas. Advantageously, the improved forest inventory system models tree count, height, and parameters to characterize the forest using optical data, synthetic-aperture radar (SAR) data, topographical data, and other data. The system, method, apparatus, and computer-readable medium described herein provide a technical improvement to modeling forests.
[0005] In accordance with some embodiments of the present disclosure, a method is disclosed. The method may include receiving, by one or more processors, an image depicting an overhead view of a wooded area, the image comprising a plurality of pixels; receiving, by the one or more processors, a set of climate data for a geographic region in which the wooded area is located; receiving, by the one or more processors, a point cloud of a digital surface model of the wooded area; concatenating, by the one or more processors, data corresponding to the plurality of pixels of the image, the set of climate data, and the point cloud into a feature vector; executing, by the one or more processors, a machine learning model using the feature vector to generate timber data for each of the plurality of pixels of the image; and generating, by the one or more processors, an interactive overlay from the timber data, the interactive overlay comprising the generated timber data for each of the plurality of pixels of the image.--, in [0003]-[0005]; and, --getting ground resources to the right place at the right time to maximize impact. --, in [0021]; --the use of satellite imagery and artificial intelligence (AI) processing techniques to remotely provide a view of an entire forest inventory across vast geographic areas and to analyze disturbance events that threaten its value. This solution helps manage inventory, carbon stock, fire damage, pest, and disease, brushing, and mill optimization.
[0025] Advantageously, the embodiments described herein track the full forest lifecycle across seasons, fusing satellite and multiple data feeds with advanced AI….Digital surface model data can be included in the list of predictors to further increase the accuracy of the model output. These sources are used to generate inputs to a model. The inputs can be SAR indices, spectral indices, and values for topographic variables.
[0028] The model may generate species distribution (e.g., the distribution and intermixing of different tree species within a forest) and/or tree mensuration (e.g., estimates of the total volume and/or biomass of available timber in forest areas and additionally models of total tree count, height and/or the diameter at breast height (DBH) parameters) data.--, in [0025]-[0028]; further see Jha: e.g., -- a computer-implemented method for generating land use land cover (LULC) classification of a geographic area is described. The computer-implemented method comprises receiving a first input defining a geographic area and a first time frame. The computer-implemented method further comprises automatically retrieving a first set of satellite images corresponding to the geographic area and the first time frame. The computer-implemented method further comprises automatically classifying the first set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting the LULC classification of the geographic area.
[0007] According to an example, the plurality of land use land cover (LULC) classes may include at least one of vegetation cover, surface water cover, built-up area, barren/open land, and cropland.
[0008] According to an example, the computer-implemented method may further comprise creating a training set including a plurality of satellite images, and automatically training a deep learning model using the training set and a neural network to develop the trained deep learning model.
[0009] According to an example, creating a training set may further comprise automatically retrieving a plurality of satellite images corresponding to a plurality of geographic areas, automatically fetching a plurality of spectral bands corresponding to the plurality of satellite images, automatically processing the plurality of spectral bands to convert digital number of each pixel of the plurality of spectral bands into reflectance or radiance values, and creating the training set in the form of creating pixel-wise shapefiles corresponding to each of the plurality of LULC classes.--, in [0006]-[0009]; and, -- the computer-implemented method may further comprise receiving a second input defining a second time frame, automatically retrieving a second set of satellite images corresponding to the geographic area and the second time frame, automatically classifying the second set of satellite images into a plurality of land use land cover (LULC) classes using a trained deep learning model, and automatically presenting a visualization depicting a comparison of the land use land cover (LULC) classes of the first and the second set of satellite images, the comparison illustrating a quantitative relative change in the land use land cover (LULC) classes of the geographic area over a time duration from the first time frame to the second time frame.
[0017] According to another exemplary embodiment, a system for generating land use land cover (LULC) classification of a geographic area is described. The system comprises at least one processor and at least one computer readable memory coupled to the at least one processor, and the processor is configured to perform all or some steps of the method described above.
[0018] According to another exemplary embodiment, a non-transitory computer readable medium is described. The non-transitory computer readable medium comprises a computer-readable code comprising instructions, which when executed by a processor, causes the processor to perform all or some steps of the method described above.
[0019] It is an object of the invention to provide a Geo-spatial artificial intelligence (Geo-AI) based fully automated computer-based method and system for predicting land use land cover (LULC) classification of a geographic area using a trained deep learning model--, in [0016]-[0019]).
Re Claims 31-33, claims 31-33 are the corresponding system claim to claims 1-3 respectively. Thus, claims 31-33 are rejected for reasons similar to those discussed in regard to claims 1-3. Further, Gibson as modified by Veronesi and Jha further disclose system of training a machine learning model to identify an expected distribution of mapping categories for a geographic area at a given time, comprising: processing circuitry configured to perform the functions (see Gibson: e.g., -- computer-implemented methods and systems are provided for pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system. To do this, the amount that each class is contained within each area image relative to the whole dataset is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken-, in abstract, and, -- computer-implemented methods and systems for generating training data for training a machine learning system to automatically extract topographic features. In particular, aspects described herein relate to a method of pre-processing image data to generate improved training data.--, in [0001], and, -- pre-processing image data to generate improved training data for training a machine learning system to automatically classify imagery. One or more images of a geographic region are captured and processed to obtain a plurality of labelled samples (also referred to as patches) for training the machine learning system, with each image containing a part of the geographic region of interest. To do this, the amount that each class is contained within each area image relative to the whole dataset (i.e., the whole geographic region captured) is measured, with the distribution of classes being weighted according to this measurement and normalised to determine how many patches per class and per area should be taken. This ensures that the distribution of patches taken from each area prioritises features that are less common….. a computer-implemented method of processing image data for use in training a machine learning system for classifying image data, the method comprising obtaining image data comprising a plurality of images, each image corresponding to a respective geographic area, processing the image data to identify one or more classes of topographic feature contained within each image, wherein processing the image data comprises determining a quantity of each respective topographic feature class contained within each image, generating a first dataset comprising a set of values for each image, wherein each value is representative of the quantity of one of the plurality of topographic feature classes contained within the respective image, and processing the first dataset to determine a number of samples required from each image for each topographic feature class. Processing the first dataset comprises (i) generating a normalised dataset, wherein the values of the first dataset are normalised across each image and each topographic feature class, and (ii) for each topographic feature class, calculating the number of samples required from each image based on the normalised dataset and a target number of samples for the respective topographic feature class. --, in [0008]-[0010]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to WEIWEN VERONESI whose telephone number is (571)270-5670. The examiner can normally be reached on Monday-Friday 8:30am-4:30pm east.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Amandeep Saini can be reached on 571-272-3382. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/WEI WEN VERONESI/Primary Examiner, Art Unit 2662