DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA . The rejections from the Office Action of 9/22/2025 are hereby withdrawn. New grounds for rejection are presented below.
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 2/20/2026 has been entered.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
Claim(s) 1, 2, 3, 5-9, 12-14, and 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Shendryk et al., Integrating satellite imagery and environmental data to predict field-level cane and sugar yields in Australia using machine learning, Elsevier, 2020 [hereinafter “Shendryk”](supplied by the Applicant on 3/2/2023); Birla, Single Image Super Resolution Using GANS, Medium.com, 11.8.2018 (supplied by the Applicant on 3/2/2023); Badrinarayanan et al., SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation, arXiv, 2016 [hereinafter “Badrinarayanan”]; Roman et al., Noise Estimation for Generative Diffusion Models, arXiv, 9.12.2021 [hereinafter “Roman”]; and Ellinger, Understanding Spatial Resolution with Drones, TLT Photography, 2017.
Regarding Claims 1 and 12, Shendryk discloses a computer-implemented method/method/instructions for use in processing image data associated with crop-bearing fields [Title – “Integrating satellite imagery and environmental data to predict field-level cane and sugar yields in Australia using machine learning”], the method comprising:
accessing, by a computing device, a first data set, the first data set including images associated with one or more fields, the images being satellite images having a spatial resolution of about one meter or more per pixel [Page 3, first column – “The satellite imagery consisted of Sentinel-1 SAR and Sentinel-2 multispectral imagery covering the area of the Wet Tropics of Australia (Fig. 1). Sentinel-1 imagery had a spatial resolution of 10 m and temporal resolution of up to 12 days, while Sentinel-2 imagery had a spatial resolution of 10−20 m and temporal resolution of up to 5 days.”Page 6, second column – “In this study, to utilize the power of OBIA approaches without relying on segmentation algorithms to delineate individual fields we trained machine learning models using predictor variables calculated within each sugarcane field (> 0.64 ha) (see Table 4) and used them for inference using predictor variables calculated within 8 × 8 pixels windows (0.64 ha) to match the smallest size of fields used in this study. Prediction maps of cane yield, CCS, sugar yield and crop variety were generated for the month of March (i.e. between January – March) for each growing season between 2016 – 2020. Then, cane yield, CCS and sugar yield prediction maps over the whole of Wet Tropics at 10 m resolution were averaged (excluding fallow crop areas identified using a machine learning model) across four mill areas (i.e. Mulgrave, South Johnstone, Tully and Macknade, see Section 2.6) and compared against actual values of cane yield, CCS and sugar yield reported by the mills in 2016, 2017 and 2018 growing seasons (Canegrowers, 2019).”].
Shendryk fails to disclose training a diffusion model architecture of a computing device based on training data, the training data including satellite images and drone images, where the satellite images have a first spatial resolution of about one meter or more per pixel, and where the drone images have a second spatial resolution of about X centimeters per pixel, where X is less than about 5 centimeters; and generating, by the computing device, based on the diffusion model, defined resolution images of the one or more fields from the first data set, the defined resolution images each having the second spatial resolution of about X centimeters per pixel, wherein the diffusion model architecture includes an encoder, a decoder, and a semantic latent space coupled between the encoder and the decoder; and wherein training the diffusion model includes: generating, by the encoder of the diffusion model architecture, a representation of each of the satellite images within the semantic latent space; and for each of the satellite images: adding, by the decoder of the diffusion model architecture, noise to the representation of the satellite image within the semantic latent space at each timestep of multiple timesteps, wherein each timestep corresponds to a particular noise level, which increases with each increase in the timesteps; and removing, by the decoder of the diffusion model architecture, the noise added to the representation of each of the satellite images within the semantic latent space at each timestep of the multiple timesteps, whereby the diffusion model is trained at each timestep and is operable to reverse the noise added to the representation of the satellite image to achieve the representation of the satellite image initially generated within the semantic latent space by the encoder.
However, Birla discloses a super-resolution GAN [Page 1 – “Image super resolution can be defined as increasing the size of small images while keeping the drop in quality to minimum, or restoring high resolution images from rich details obtained from low resolution images.”] through which a generator uses low resolution images to generate high resolution images [See Page 4]. Birla discloses doing so through training a model using appropriate training data [See Page 2]. It would have been obvious to use a super-resolution GAN in order to obtain higher resolution images from the satellites of Shendryk for use in analyzing crop-bearing fields.
Badrinarayanan discloses model architecture that includes an encoder, a decoder, and a semantic latent space coupled between the encoder and the decoder; and generating, by the encoder of the diffusion model architecture, a representation of each of the images within the semantic latent space [Abstract – “We present a novel and practical deep fully convolutional neural network architecture for semantic pixel-wise segmentation termed SegNet. This core trainable segmentation engine consists of an encoder network, a corresponding decoder network followed by a pixel-wise classification layer. The architecture of the encoder network is topologically identical to the 13 convolutional layers in the VGG16 network [1]. The role of the decoder network is to map the low resolution encoder feature maps to full input resolution feature maps for pixel-wise classification. The novelty of SegNet lies is in the manner in which the decoder upsamples its lower resolution input feature map(s). Specifically, the decoder uses pooling indices computed in the max-pooling step of the corresponding encoder to perform non-linear upsampling. This eliminates the need for learning to upsample. The upsampled maps are sparse and are then convolved with trainable filters to produce dense feature maps.”See Fig. 1 – “Convolutional Encoder-Decoder.” The “semantic latent space” present between the encoder/decoder as they are used for semantic (“semantic”) segmentation of low resolution feature maps (“latent space”).]. It would have been obvious to use such architecture in generating higher resolution images because Badrinarayanan discloses that this is a contemplated use [Page 4, first column – “Other applications where pixel wise predictions are made using deep networks are image super-resolution”].
Roman discloses the use of a diffusion model that defines a forward diffusion process and a reverse diffusion process to reverse the diffusion of noise in the noisy sample(s) [Page 1, first column – “An emerging class of non-autoregessive models is the one of Denoising Diffusion Probabilistic Models (DDPM). Such methods use diffusion models and denoising score matching in order to generate images (Ho, Jain, and Abbeel 2020) and speech (Chen et al. 2020). The DDPM model learns to perform a diffusion process on a Markov chain of latent variables. The diffusion process transforms a data sample into Gaussian noise. During inference the reverse process is used, which is called the denoising process.”] and that such a diffusion model can be one adding, by the of the diffusion model architecture, noise to the representation at each timestep of multiple timesteps, wherein each timestep corresponds to a particular noise level, which increases with each increase in the timesteps; and removing, by the diffusion model architecture, the noise added at each timestep of the multiple timesteps, whereby the diffusion model is trained at each timestep and is operable to reverse the noise added to achieve the representation initially generated [Page 2, first column – “The recent TimeGrad model (Rasul et al. 2021) is a diffusion process for probabilistic time series forecasting, which was shown empirically to outperform Transformers (Vaswani et al. 2017) and LSTMs (Hochreiter and Schmidhuber 1997) on some datasets.”
Pages 2-3 – “Denoising Diffusion Probabilistic Model (DDPM) are neural network that learn the gradients of the data log density …The formalization of Denoising Diffusion Probabilistic Models (DDPM) by Ho et al(Ho, Jain, and Abbeel 2020) employs a parameterized Markov chain trained using variational inference, in order to produce samples matching the data after finite time. The transitions of this chain are learned to reverse a diffusion process. This diffusion process is defined by a Markov chain that gradually adds noise in the data with
a noise schedule β1,...βN …At each iteration, the diffusion process adds Gaussian noise, according to the noise schedule …One can use this to implement the DDPM training algorithm …One can use this update rule in Eq. 9 to sample from the data distribution, by starting from a Gaussian noise and then step-by-step reversing the diffusion process.”]
It would have been obvious to use such a diffusion model and diffusion model techniques in the image generation of Birla in order to achieve denoising of the low resolution images.
Ellinger discloses that an aerial drone can be used to image at such a resolution [Page 4 – “If you used the Pix4D GSD Calculator and you wanted to obtain a 3cm/pixel spatial resolution you would fly at an altitude of 109 meters (358ft). … Yes, you can achieve 1cm spatial resolution by drone[.]”]. It would have been obvious to use an aerial camera to produce such high resolution images such that they could be used as the high resolution images needed by Birla in order to use the SRGAN to enhance the lower resolution satellite images.
Shendryk, as modified, would disclose deriving, by the computing device, index values for the one or more fields into a combined metric [Page 3, second column – “In contrast to previous studies that exclusively relied on NDVI (Begue et al., 2010; Morel et al., 2014) or green NDVI (GNDVI) (Rahman and Robson, 2020) for yield prediction, in this study we derived 45 normalized difference spectral indices (NDSIs) from Sentinel-2 imagery, that previously showed to be useful in differentiating vegetation types (Shendryk et al., 2020b). NDSIs were calculated in succession from Blue to SWIR-2 (Table 3) spectral bands as follows:
NDSI(i,j) = (Ri − Rj)/(Ri + Rj),
where R is the spectral reflectance, and i and j are numbers indicating the wavelengths (nm). Throughout this paper each NDSI is denoted as a combination of three-letter acronyms in Table 3 (e.g. NDSI(RED,NI1) is denoted as REDNI1 while NDSI(GRE,NI1) is denoted as GRENI1 and is an inverse of GNDVI).”Page 7, second column – “Sentinel-2 derived NDSIs and DEM derived predictors were the strongest in predicting cane yield, CCS and sugar yield.”], based on the defined resolution images of the one or more fields [per Birla];
aggregating, by the computing device, the index values for the one or more fields with at least one environmental metric for the one or more fields into a combined metric for the one or more fields [Page 4, first column – “Climate data were downloaded from the database of Australian climate data (SILO, 2020) for 98 stations spread across the Wet Tropics of Australia. These data contained daily averages for (1) solar radiation – total incoming downward shortwave radiation on a horizontal surface in MJ/m2 (RAD), (2) maximum temperature in ºC (TMAX), (3) minimum temperature ºC (TMIN), (4) vapour pressure in hPA (VP) as well as daily totals of (5) evaporation in mm (EVAP) and (6) rainfall in mm (RAIN).”Page 5, second column – “In total, there were 371 predictor variables, extracted from Sentinel1 (21 predictors, i.e. 7 statistics (Table 4) extracted from 3 SAR bands), Sentinel-2 (315 predictors, i.e. 7 statistics (Table 4) exracted from 45 NDSIs), DEM (28 predictors, i.e. 7 statistics (Table 4) extracted from 4 DEM variables), soil (1 predictor) and climate (6 predictors, i.e. average (avg) value of 6 climate variables) mosaics.”See Table 6 – All predictors.];
encoding the combined metric to the defined resolution images [per Birla] of the one or more fields [Title – “Integrating satellite imagery and environmental data to predict field-level cane and sugar yields in Australia using machine learning”];
predicting, by the computing device, a plot yield for the one or more fields, based on the combined metric [See section 2.8. Yield prediction, particularly – “In total, there were 371 predictor variables, extracted from Sentinel1 (21 predictors, i.e. 7 statistics (Table 4) extracted from 3 SAR bands), Sentinel-2 (315 predictors, i.e. 7 statistics (Table 4) exracted from 45 NDSIs), DEM (28 predictors, i.e. 7 statistics (Table 4) extracted from 4 DEM variables), soil (1 predictor) and climate (6 predictors, i.e. average (avg) value of 6 climate variables) mosaics.”See section 3.2. Yield predictionsSee Table 6 – All predictors.]; and
storing, by the computing device, the predicted plot yield for the one or more fields [See Fig. 8] in a memory [Inherent given the use of machine learning. See section 2.8.3. Machine learning inference].
Regarding Claims 2 and 13, Shendryk discloses that the first data set includes satellite images of the one or more fields, in which a crop is grown [Title – “Integrating satellite imagery and environmental data to predict field-level cane and sugar yields in Australia using machine learning”], but the combination fails to disclose that X is less than or equal to about 1 centimeter.
However, Ellinger discloses that an aerial drone can be used to image at such a resolution [Page 4 – “If you used the Pix4D GSD Calculator and you wanted to obtain a 3cm/pixel spatial resolution you would fly at an altitude of 109 meters (358ft). … Yes, you can achieve 1cm spatial resolution by drone[.]”]. It would have been obvious to use an aerial camera to produce such high resolution images such that they could be used as the high resolution images needed by Birla in order to use the SRGAN to enhance the lower resolution satellite images.
Regarding Claims 3 and 14, Roman discloses the use of a denoising diffusion model [Page 1, first column – “An emerging class of non-autoregessive models is the one of Denoising Diffusion Probabilistic Models (DDPM). Such methods use diffusion models and denoising score matching in order to generate images (Ho, Jain, and Abbeel 2020) and speech (Chen et al. 2020). The DDPM model learns to perform a diffusion process on a Markov chain of latent variables. The diffusion process transforms a data sample into Gaussian noise. During inference the reverse process is used, which is called the denoising process.”], wherein each of the timesteps is associated with a certain noise level [Page 2, first column – “The recent TimeGrad model (Rasul et al. 2021) is a diffusion process for probabilistic time series forecasting, which was shown empirically to outperform Transformers (Vaswani et al. 2017) and LSTMs (Hochreiter and Schmidhuber 1997) on some datasets.”Page 3, first column – “One can use this update rule in Eq. 9 to sample from the data distribution, by starting from a Gaussian noise and then step-by-step reversing the diffusion process.”].
Regarding Claim 5, Shendryk discloses defining, by the computing device, a field level image for each of the one of more fields, from the defined resolution images [See Figs. 2 and 8]; and
wherein deriving the index values for the one or more fields includes deriving the index values for each of the field level images for the one or more fields [Page 3, second column – “In contrast to previous studies that exclusively relied on NDVI (Begue et al., 2010; Morel et al., 2014) or green NDVI (GNDVI) (Rahman and Robson, 2020) for yield prediction, in this study we derived 45 normalized difference spectral indices (NDSIs) from Sentinel-2 imagery, that previously showed to be useful in differentiating vegetation types (Shendryk et al., 2020b). NDSIs were calculated in succession from Blue to SWIR-2 (Table 3) spectral bands as follows:
NDSI(i,j) = (Ri − Rj)/(Ri + Rj),
where R is the spectral reflectance, and i and j are numbers indicating the wavelengths (nm). Throughout this paper each NDSI is denoted as a combination of three-letter acronyms in Table 3 (e.g. NDSI(RED,NI1) is denoted as REDNI1 while NDSI(GRE,NI1) is denoted as GRENI1 and is an inverse of GNDVI).”Page 7, second column – “Sentinel-2 derived NDSIs and DEM derived predictors were the strongest in predicting cane yield, CCS and sugar yield.”].
Regarding Claim 6, Shendryk discloses that deriving the index values includes deriving each of the index values based on the following:
index value = (nir — red) / (nir +red); and
wherein nir is a near infrared band value of each of the field level images and red is a red band value of each of the field level images [Page 3, second column – “In contrast to previous studies that exclusively relied on NDVI (Begue et al., 2010; Morel et al., 2014) or green NDVI (GNDVI) (Rahman and Robson, 2020) for yield prediction, in this study we derived 45 normalized difference spectral indices (NDSIs) from Sentinel-2 imagery, that previously showed to be useful in differentiating vegetation types (Shendryk et al., 2020b). NDSIs were calculated in succession from Blue to SWIR-2 (Table 3) spectral bands as follows:
NDSI(i,j) = (Ri − Rj)/(Ri + Rj),
where R is the spectral reflectance, and i and j are numbers indicating the wavelengths (nm). Throughout this paper each NDSI is denoted as a combination of three-letter acronyms in Table 3 (e.g. NDSI(RED,NI1) is denoted as REDNI1 while NDSI(GRE,NI1) is denoted as GRENI1 and is an inverse of GNDVI).”Page 7, second column – “Sentinel-2 derived NDSIs and DEM derived predictors were the strongest in predicting cane yield, CCS and sugar yield.”].
Regarding Claim 7, Shendryk discloses that the index values quantifyf a vegetation greenness of the one or more fields [Page 3, second column – “In contrast to previous studies that exclusively relied on NDVI (Begue et al., 2010; Morel et al., 2014) or green NDVI (GNDVI) (Rahman and Robson, 2020) for yield prediction, in this study we derived 45 normalized difference spectral indices (NDSIs) from Sentinel-2 imagery, that previously showed to be useful in differentiating vegetation types (Shendryk et al., 2020b). NDSIs were calculated in succession from Blue to SWIR-2 (Table 3) spectral bands as follows:
NDSI(i,j) = (Ri − Rj)/(Ri + Rj),
where R is the spectral reflectance, and i and j are numbers indicating the wavelengths (nm). Throughout this paper each NDSI is denoted as a combination of three-letter acronyms in Table 3 (e.g. NDSI(RED,NI1) is denoted as REDNI1 while NDSI(GRE,NI1) is denoted as GRENI1 and is an inverse of GNDVI).”Page 7, second column – “Sentinel-2 derived NDSIs and DEM derived predictors were the strongest in predicting cane yield, CCS and sugar yield.”].
Regarding Claims 8 and 16, the combination would disclose that the images of the first data set further include a temporal resolution of one image per N number of days, where N is an integer less than about 30 [Page 3, first column of Shendryk – “The satellite imagery consisted of Sentinel-1 SAR and Sentinel-2 multispectral imagery covering the area of the Wet Tropics of Australia (Fig. 1). Sentinel-1 imagery had a spatial resolution of 10 m and temporal resolution of up to 12 days, while Sentinel-2 imagery had a spatial resolution of 10−20 m and temporal resolution of up to 5 days.”]; and
that the defined resolution images of the one or more fields include said temporal resolution [Super resolution enhancement of the satellite images per Birla].
Regarding Claim 9, Shendryk discloses the at least one environmental metric includes at least one of: precipitation, solar radiation, and/or temperature [Page 4, first column – “Climate data were downloaded from the database of Australian climate data (SILO, 2020) for 98 stations spread across the Wet Tropics of Australia. These data contained daily averages for (1) solar radiation – total incoming downward shortwave radiation on a horizontal surface in MJ/m2 (RAD), (2) maximum temperature in ºC (TMAX), (3) minimum temperature ºC (TMIN), (4) vapour pressure in hPA (VP) as well as daily totals of (5) evaporation in mm (EVAP) and (6) rainfall in mm (RAIN).”].
Claim(s) 10 and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Shendryk et al., Integrating satellite imagery and environmental data to predict field-level cane and sugar yields in Australia using machine learning, Elsevier, 2020 [hereinafter “Shendryk”](supplied by the Applicant on 3/2/2023); Birla, Single Image Super Resolution Using GANS, Medium.com, 11.8.2018 (supplied by the Applicant on 3/2/2023); Badrinarayanan et al., SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation, arXiv, 2016 [hereinafter “Badrinarayanan”]; Roman et al., Noise Estimation for Generative Diffusion Models, arXiv, 9.12.2021 [hereinafter “Roman”]; Ellinger, Understanding Spatial Resolution with Drones, TLT Photography, 2017; and Guo et al. (US 20220217894 A1)[hereinafter “Guo”].
Regarding Claims 10 and 17, Shendryk discloses that aggregating the index values for the one or more fields with the at least one environmental metric for the one or more fields includes aggregating the index values with the at least one environmental metric [Page 5, second column – “In total, there were 371 predictor variables, extracted from Sentinel1 (21 predictors, i.e. 7 statistics (Table 4) extracted from 3 SAR bands), Sentinel-2 (315 predictors, i.e. 7 statistics (Table 4) exracted from 45 NDSIs), DEM (28 predictors, i.e. 7 statistics (Table 4) extracted from 4 DEM variables), soil (1 predictor) and climate (6 predictors, i.e. average (avg) value of 6 climate variables) mosaics.”See Table 6 – All predictors.], but fails to disclose aggregating the value/metric, by using one of inverted variance weighting and convolutional neural networking, into a combined metric; and wherein predicting the plot yield is based on the combined metric.
However, Guo discloses aggregating overhead farm images and environmental data through use of a neural network in the determination of an SOC metric [Paragraph [0025] – “As shown in FIG. 1 and mentioned previously, some farm vehicles may be operated at least partially autonomously, and may include, for instance, unmanned aerial vehicle 107.sub.1 that carries a vision sensor 108.sub.1 that acquires vision sensor data such as digital images from overhead field(s) 112.”
Paragraph [0027] – “Local data module 116 may be configured to gather, collect, request, obtain, and/or retrieve ground truth observational data from a variety of different sources, such as agricultural personnel and sensors and software implemented on robot(s), aerial drones, and so forth. Local data module 116 may store that ground truth observational data in one or more of the databases 115, 121, or in another database (not depicted). This ground truth observational data may be associated with individual agricultural fields or particular positional coordinates within such field(s), and may include various types of information derived from user input and sensor output related to soil composition (e.g., soil aeration, moisture, organic carbon content, etc.), agricultural management practices (e.g., crop plantings, crop identification, crop rotation, irrigation, tillage practices, etc.), terrain (e.g., land elevation, slope, erosion, etc.), climate or weather (e.g., precipitation levels/frequency, temperatures, sunlight exposure, wind, humidity, etc.), and any other features, occurrences, or practices that could affect the agricultural conditions of the field(s) and which could be identified based on analyzing sensor output and/or user input and/or generated based on such identified data.”
Paragraph [0036] – “SOC inference module 128 can receive, gather, or otherwise obtain the digital images, the operational data, and observational data in order to use the types of data to generate predicted SOC measurements for the field(s) 112.”
Paragraph [0040] – “SOC inference module 128 may take the form of recurrent neural network(s) (“RNN”), the aforementioned CNNs, long short-term memory (“LSTM”) neural network(s), gated recurrent unit (“GRU”) recurrent network(s), feed forward neural network(s), or other types of memory networks.”]. It would have been obvious to determine/aggregate such a metric using the index values and environmental metric and use it in predicting crop yield because Guo discloses that SOC is relevant to crop yield [Paragraph [0059] – “In some implementations, at block 408, crop yield may be predicted as well (crop yield may be correlated with SOC extracted from and/or added to the soil).”].
Response to Arguments
Applicant argues:
PNG
media_image1.png
700
972
media_image1.png
Greyscale
PNG
media_image2.png
164
968
media_image2.png
Greyscale
PNG
media_image3.png
288
977
media_image3.png
Greyscale
Examiner’s Response:
Applicant’s argument is convincing. The rejections under 35 USC 101 are hereby withdrawn.
Applicant argues:
PNG
media_image4.png
422
969
media_image4.png
Greyscale
PNG
media_image5.png
504
974
media_image5.png
Greyscale
PNG
media_image6.png
249
972
media_image6.png
Greyscale
Examiner’s Response:
The Examiner agrees. New grounds for rejection are presented above.
Applicant argues:
PNG
media_image7.png
502
974
media_image7.png
Greyscale
Examiner’s Response:
The Examiner respectfully disagrees. Although Birla discloses the use of a GAN in image production, the generator produces the images; there is no issue with the applying the teachings of Roman relative to the generator. Paragraphs [0037] and [0052] of the instant Specification disclose that the diffusion model can be used in combination with a generator.
The Examiner agrees the combination does not disclose the amended subject matter. New grounds for rejection is presented above.
The use of AI enhanced images in combination with Shendryk would have been understood by one having ordinary skill in the art as an improvement (or, at least, potential improvement) to Shendryk; the principle of operation would not be changed but, rather, enhanced.
Ellinger discloses that an aerial drone can be used to image at high resolution [Page 4 – “If you used the Pix4D GSD Calculator and you wanted to obtain a 3cm/pixel spatial resolution you would fly at an altitude of 109 meters (358ft). … Yes, you can achieve 1cm spatial resolution by drone[.]”]. It would have been obvious to use an aerial camera to produce such high resolution images such that they could be used as the high resolution images needed by Birla in order to use the SRGAN to enhance the lower resolution satellite images of Shendryk.
Applicant argues:
PNG
media_image8.png
245
976
media_image8.png
Greyscale
Examiner’s Response:
The Examiner respectfully disagrees. Shendryk, as modified, would disclose encoding the combined metric to the defined resolution images [per Birla] of the one or more fields [Title – “Integrating satellite imagery and environmental data to predict field-level cane and sugar yields in Australia using machine learning”].
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Gandikota et al., RTC-GAN: REAL-TIME CLASSIFICATION OF SATELLITE IMAGERY USING DEEP GENERATIVE ADVERSARIAL NETWORKS WITH INFUSED SPECTRAL INFORMATION, IEEE, 2020
Jiang et al., GAN-BASED MULTI-LEVEL MAPPING NETWORK FOR SATELLITE IMAGERY SUPER-RESOLUTION, IEEE, 2019
Liu et al., PSGAN: A GENERATIVE ADVERSARIAL NETWORK FOR REMOTE SENSING IMAGE PAN-SHARPENING, IEEE, 2018
US 20200125929 A1 – CROP YIELD PREDICTION AT FIELD-LEVEL AND PIXEL-LEVEL
US 20220335715 A1 – PREDICTING VISIBLE/INFRARED BAND IMAGES USING RADAR REFLECTANCE/BACKSCATTER IMAGES OF A TERRESTRIAL REGION
US 20210012109 A1 – SYSTEM AND METHOD FOR ORCHARD RECOGNITION ON GEOGRAPHIC AREA
US 20180189564 A1 – METHOD AND SYSTEM FOR CROP TYPE IDENTIFICATION USING SATELLITE OBSERVATION AND WEATHER DATA
US 20220198221 A1 – ARTIFICIAL INTELLIGENCE GENERATED SYNTHETIC IMAGE DATA FOR USE WITH MACHINE LANGUAGE MODELS
US 20180211156 A1 – CROP YIELD ESTIMATION USING AGRONOMIC NEURAL NETWORK
US 20210201024 A1 – CROP IDENTIFICATION METHOD AND COMPUTING DEVICE
US 20210397836 A1 – USING EMPIRICAL EVIDENCE TO GENERATE SYNTHETIC TRAINING DATA FOR PLANT DETECTION
US 20210312591 A1 – SYSTEMS AND METHOD OF TRAINING NETWORKS FOR REAL-WORLD SUPER RESOLUTION WITH UNKNOWN DEGRADATIONS
Nichol et al., GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models, arXiv, 3.8.2022
Song et al., DENOISING DIFFUSION IMPLICIT MODELS, arXiv, 2020
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KYLE ROBERT QUIGLEY whose telephone number is (313)446-4879. The examiner can normally be reached 9AM-5PM EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Arleen Vazquez can be reached at (571) 272-2619. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/KYLE R QUIGLEY/Primary Examiner, Art Unit 2857
Prosecution Insights