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 .
Response to Amendment
The amendments filed 01/28/2026 have been entered. Claims 1, 3-9 remain pending in the application.
Applicant’s amendment, with respect to the claim rejection(s) of claim 1, 3-9 under 35 U.S.C 101 filed 10/28/2025 have been considered and are persuasive. Therefore, the previous rejections as set forth in the previous office action has been removed.
Applicant’s amendment, with respect to the claim rejection(s) of claim 1, 3-9 under 35 U.S.C 103 filed 10/28/2025 have been considered and are not persuasive. Therefore, the previous rejections as set forth in the previous office action will be maintained.
The Applicant argues that amended independent claim 1 now requires obtaining, for each of the N pieces of input data, a reliability degree with respected to the classified class by multiplying, as a function of the similarity degree, a classification output value and a positive coefficient that is not equal to zero. Applicant contends that Van der Meer only discloses a probability of spectral discrimination based on a normalization constant and spectral similarity measures, but does not teach calculating the claimed reliability degree by multiply a similarity degree, a classification output with respect to the classified class, and a positive non-zero coefficient. Applicant therefore asserts that the combination does not teach or suggest amended independent claim 1, and that the remaining independent and dependent claims are patentable for at least the same reason.
The examiner respectfully disagrees. Applicant’s argument is not persuasive. The rejection relies on the combined teaching of Pabbaraju and Van der Meer. Pabbaraju teaches classifying each pixel of an input image into one of a set of classes and providing classification output/label information for the classified pixel. Van der Meer teaches comparing an unknown imaged pixel spectrum with known spectra from a spectral library using spectral similarity measures, where the known spectra correspond to the claimed known feature spectrum group and the unknown pixel spectrum corresponds to the claimed feature spectrum for each input pixel. Van der Meer further teaches calculating a probability of spectral discrimination based on the similarity measure and a normalization/scaling factor at page 4 section 2 equation 7, where the probability indicates how reliably the feature spectrum is predicted to belong to a spectral class. A person ordinary skill in the art would have understood that normalizing by the normalization constant is equivalent to applying a scaling factor, and that the scaling factor must be positive/non-zero for the probability calculation to be valid. Thus, the normalization/scaling factor corresponds to the amended claimed coefficient that is not equal to zero. Furthermore, a person of ordinary skill in the art would have been motivated to apply Van der Meer’s similarity-based normalization to Pabbaraju’s classification output value for the classified pixel, so that the output is weighted by how well the pixel’s feature spectrum matches the known spectral library, because applying a weight or scaling factor to an output is commonly implemented by multiplying the output by that weight or scaling factor. Thus, the resulting probability/reliability degree is obtained as a function of the Van der Meer’s similarity degree, Pabbaraju’s classification output value, and Van der Meer’s normalization/scaling factor, thereby suggesting obtaining the reliability degree with respect to the classified class by multiplying as a function of the similarity degree, a classification output value, and a positive coefficient, as claimed. Therefore, the teaching combination by Pabbaraju in view of Van der Meer still teaches or at least suggests the amended claim.
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 1-3, 5, 7-9 are rejected under 35 U.S.C. 103 as being unpatentable over Pabbaraju et.al (US 20230107917 A1), in view of Van der Meer et.al (NPL: The effectiveness of spectral similarity measures for the analysis of hyperspectral imagery)
Regarding claim 1,
Pabbaraju teaches the preamble “A method of executing class classification processing relating to M classes using a machine learning model including a vector neural network including a plurality of vector neuron layers, where M is an integer equal to or greater than 2” (paragraph 57 “Classifier 514 may be configured to classify input signals x into one or more labels using a machine learning (ML) algorithm, such as a neural network described above”, paragraph 20 “The inputs may also include C as a number of classes or a set of classes”, and paragraph 47 “An image semantic segmentation model ... may seek solutions to the following problem: given an image 301 x ... and set of classes C, classify each pixel in x to a class in C”. Pabbaraju discloses a method of classifying an image using a machine learning model classifier with machine learning algorithm configured as a neural network with layers. The classifier can classify input data into classes, wherein there is a set of classes C, suggesting that there is more than one class for the input data to be classified, which is analogous to M classes equal to or greater than two within the claim.)
Pabbaraju teaches the limitation “(a) capturing, via a camera, a plurality of images of a target object” (paragraph 45 “As an example, the raw source data 215 may include raw video images from a camera.” Pabbaraju discloses raw input images, wherein the images are obtained from a camera, which correspond to the plurality of images captured via the camera, as claimed.)
Pabbaraju teaches the limitation “(b) generating N pieces of input data from one target object, where N is an integer equal to or greater than 2, and the N pieces of input data includes at least one image of the captured plurality of images” (paragraph 47 “An image semantic segmentation model ... may seek solutions to the following problem: given an image 301 x ... and set of classes C, classify each pixel in x to a class in C”, paragraph 48 “The system may include a super pixel algorithm 307. Super pixel algorithms may group neighboring pixels with similar pixel values to one super pixels” Pabbaraju discloses one or more images input, wherein the image is configured with various pixel to be classified. Some of the pixel may be grouped together to form one or more super pixel for classification. Furthermore, those pixels are obtained from and included in the received image, wherein the image is obtained from the camera as disclosed above. Accordingly, the pixels correspond to the claimed N pieces of input data, wherein N is equal to or greater than two, and the received image containing those pixels corresponds to the claimed at least one image of the captured plurality of images. Thus, Pabbaraju teaches or at least suggests the limitation.)
Pabbaraju teaches a part of the limitation “(c) inputting each of the N pieces of input data to the machine learning model, and obtaining, for each of the N pieces of input data, M classification output values that are output from an output layer of the machine learning model, a classified class, and a feature ... that is obtained from an output of a specific layer of the machine learning model;” (paragraph 34 “The system 100 may further ... provide an iterative function as a substitute for a stack of layers of the neural network to be trained. Here, respective layers of the stack of layers being substituted may have mutually shared weights and may receive, as input, an output of a previous layer, or for a first layer of the stack of layers, an initial activation, and a part of the input of the stack of layers”, paragraph 47 “An image semantic segmentation model ... may seek solutions to the following problem: given an image 301 x ... and set of classes C, classify each pixel in x to a class in C”, paragraph 48 “The system may include a super pixel algorithm 307. Super pixel algorithms may group neighboring pixels with similar pixel values to one super pixels ... The system may calculate the super pixel loss 309 based on sampling based on a majority vote”, paragraph 42 “The system 200 may implement a machine-learning algorithm 210 that is configured to analyze the raw source dataset 215. The raw source dataset 215 may include raw or unprocessed sensor data that may be representative of an input dataset for a machine-learning system”, and paragraph 45 “The machine-learning algorithm 210 may be configured to identify a particular feature in the raw source data 215.” Pabbaraju discloses the classification procedure of the neural network classifier model. The classifier model may be configured as a neural network with input/output layers as suggested by paragraph 34. The classifier can classify each pixel (N pieces of input data within the claim) in input image x to a class (one classified class within the claim) in C, wherein the classified pixel may belong to a super pixel within the image, and the classification can be follow-up with a calculated super pixel loss for each super pixel, which is analogous to M classification output values within the claim. Pabbaraju further discloses the system of the machine-learning algorithm may be configured to identify a particular feature, wherein one of ordinary skilled in the art may configure the function to identify feature within data as a layer within the neural network classifier model to identify feature of the classified pixel. The identified feature may correspond to the spectrum feature as configured based on the combination with the teaching by Van der Meer with the motivation to combine the teachings below.)
Pabbaraju teaches a part of the limitation “(d) the ... feature ... being obtained from the output of the specific layer when a plurality of pieces of teaching data are input to the machine learning model” (paragraph 43 “The computer system 200 may store a training dataset 212 for the machine-learning algorithm 210. The training dataset 212 may represent a set of previously constructed data for training the machine-learning algorithm 210 ... The training dataset 212 may include a set of source data that has corresponding outcomes or results that the machine-learning algorithm 210 tries to duplicate via the learning process”, and paragraph 44 “The machine-learning algorithm 210 may be operated in a learning mode using the training dataset 212 as input. ... Since the training dataset 212 includes the expected results, the machine-learning algorithm 210 can determine when performance is acceptable. After the machine-learning algorithm 210 achieves a predetermined performance level ... the machine-learning algorithm 210 may be executed using data that is not in the training dataset 212”, and paragraph 45 “The machine-learning algorithm 210 may be configured to identify a feature in the raw source data 215 as a predetermined feature (e.g., pedestrian). The raw source data 215 may be derived from a variety of sources. For example, the raw source data 215 may be actual input data collected by a machine-learning system.” Pabbaraju discloses a process of obtaining a set of previously constructed data that has corresponding outcomes or results for training the machine-learning algorithm, which suggest a pre-labeled data to train the machine learning model. The machine learning model can be trained with such data to determine an expected performance for later execution with new data that is not in the training dataset, where in the new data may be image data with pixel as mentioned above as configured by one of ordinary skilled in the art. The machine learning algorithm and model may be the classifier neural network model to classify these constructed data. The set of previously constructed data that has corresponding outcomes or results corresponds to the teaching data within the claim, as these previously constructed data help teach the model an expected performance to be achieved. Pabbaraju further discloses identified feature, wherein one of ordinary skilled in the art may configure the function to identify feature within data as a layer within the neural network classifier model to identify feature of the classified constructed data.)
Pabbaraju teaches the limitation “(e) executing, for each of the N pieces of input data, a vote for the classified class, based on the reliability degree with respect to the classified class, and determining a class determination result for the target object, based on a result of the vote” (paragraph 21 “With respect to majority-vote Superpixel loss Lmv superpixel, for each superpixel in the image, the system may determine the most voted class in the model output across all pixels in the superpixel. For each pixel in an image, use the most voted class of the corresponding superpixel as a target”, paragraph 31 “For each image x in D, the system and method may ... compute the loss component associated with that model. The system and method may also compute the following losses: a majority-vote super pixel loss ... The majority-vote Superpixel loss may, for all pixels in the image, set the most voted class in the corresponding superpixels as the target”, and paragraph 46 “A machine-learning algorithm 210 may generate a confidence level or factor for each output generated. For example, a confidence value that exceeds a predetermined high-confidence threshold may indicate that the machine-learning algorithm 210 is confident that the identified feature corresponds to the particular feature.” Pabbaraju discloses a majority voting function within each super pixel, wherein for each super pixel in the image, the system may determine the most voted class in the model output across all pixels in the super pixel, wherein this function of majority voting for all pixel is analogous to the claimed executing a vote for the classified class for each of the N pieces of input data. Pabbaraju also discloses a confidence level or factor for each output generated, suggesting a confidence level associated with each classified class of each pixel, such that the voting may rely on this confidence level to determine if the classification is correct to finally determine a class of the image.)
Pabbaraju does not teach a part of the limitation “(c) ... a feature spectrum that is obtained ...”. However, van der Meer teaches this limitation (page 3 fig.1 “An area on the ground is imaged at high spectral resolution and spectral sampling such that after systems corrections a reflectance spectrum can be retrieved for each pixel”. van der Meer discloses methods to classify image data, which include calculating a similarity measure between known library spectra to pixel spectra of an image and to assess the performance of such similarity measure. Within the disclosure, Van der Meer discloses obtaining the pixel spectra for similarity comparison. Figure 1 demonstrate how spectra can be obtained based on hyperspectral remote sensing, which take an image of an area on the ground at high spectral resolution and spectral sampling such that a spectrum can be retrieved for each pixel for similarity comparison.)
Pabbaraju in view of van der Meer teaches or at least suggests a part the limitation “(d) obtaining a similarity degree between a known feature spectrum group and the feature spectrum for each of the N pieces of input data, the known feature spectrum group being obtained ... obtaining, for each of the N pieces of input data, a reliability degree with respect to the classified class by multiplying as a function of the similarity degree, a classification output value of the M classification output values with respect to the classified class, and a positive coefficient that is not equal to zero” (Pabbaraju discloses at paragraph 47 “An image semantic segmentation model ... may seek solutions to the following problem: given an image ... and set of classes C, classify each pixel in x to a class in C” Pabbaraju teaches classifying image pixel input into a set of classes and providing pixel classification output. Van der Meer discloses at Page 3 section 1 column 1 “The first suite of techniques strives at finding a measure of spectral similarity between a known spectrum from a spectral library with unknown imaged pixel spectra”, Page 4 section 2 column 1-2 “The performance of the following four spectral similarity measures was evaluated: (1) spectral correlation measure (SCM; Van der Meer and Bakker, 1997), (2) the spectral angle measure (SAM; Kruse et al., 1993), (3) the Euclidean distance measure (ED) and (4) the spectral information divergence (SID; Chang, 2000). In this section, these measures are briefly introduced ... The spectral correlation measure ... is calculated as the correlation coefficient of the pixel (portrayed as vector in a n dimensional feature space)”, and Page 5-6 section 3 column 1-2 equation 7 “After defining a number of spectral similarity measures, a objective statistical criteria needs to be defined to evaluate the performance of these measures. The definition of ‘optimal performance’ is intrinsically linked to the question ‘how well can the spectral measure distinguish between various spectral classes’. That implies that we are less interested in the performance of a spectral measure relative to mapping one spectral class of interest, but we seek a spectral measure that can optimally map a set of spectral classes on the basis of a set of spectral input vectors. ... The probability of spectral discrimination calculates the probability for all spectra ... In general, the higher the probability, the better is capability of the set of spectra (i.e., the spectral signature library) to predict the pixel spectrum, t. Thus, this probability tells the user that the spectral signature selected are capable and to what extent to map the selected (unknown) pixel vector”. Van der Meer discloses techniques of finding a measure of spectral similarity between a known spectrum from a spectral library with unknown or selected sampled imaged pixel spectra using some similarity measure techniques. In view of Pabbaraju’s teaching, the sampled imaged pixel spectra correspond to the feature spectrum for each of the N pieces of input data/pixels, and the known spectrum from a spectral library corresponds to the known feature spectrum group, as claimed. Accordingly, the similarity measure corresponds to the claimed similarity degree. Van der Meer further discloses obtaining a probability that a pixel t belongs to a class k based on such similarity measure. According to the equation 7, the probability is calculated based on the similarity measures for the target spectrum, t, relative to the other spectra in the spectral library sk. The probability corresponds to the claimed reliability degree because as the higher the probability, the better the capability of the set of spectra to predict the pixel spectrum t belongs to a class k. In other word, the probability indicates how confident (reliably) the pixel spectrum t is predicted as class k based on its similarity with a known spectrum from the spectral library. Van der Meer also discloses normalizing the similarity measure using a normalization constant in equation 7. A person ordinary skill in the art would have understood that normalizing by the normalization constant is equivalent to applying a scaling factor, and that the scaling factor must be positive/non-zero for the probability calculation to be valid. Thus, the normalization/scaling factor corresponds to the claimed positive coefficient that is not equal to zero. Finally, a person of ordinary skill in the art would have been motivated to apply Van der Meer’s similarity-based normalization to Pabbaraju’s classification output value for the classified pixel, so that the output is weighted by how well the pixel’s feature spectrum matches the known spectral library, because applying a weight or scaling factor to an output is commonly implemented by multiplying the output by that weight or scaling factor. Thus, the resulting probability/reliability degree is obtained as a function of the Van der Meer’s similarity degree, Pabbaraju’s classification output value, and Van der Meer’s normalization/scaling factor, thereby suggesting obtaining the reliability degree with respect to the classified class by multiplying as a function of the similarity degree, a classification output value, and a positive coefficient, as claimed.)
Before the effective filing date, it would have been obvious to a person having ordinary skill in the art to combine the teaching of a method of classifying an image with pixels into one or more classes using a machine learning model classifier with machine learning algorithm configured as a neural network with layers by Pabbaraju with the teaching of methods to classify image data, which include calculating a similarity measure between known library spectra to pixel spectra of an image and to assess the confidence of the similarity measure by van der Meer. The motivation to do so is referred to in Van der Meer’s disclosure (Page 4 section 1 “The most readily used techniques for the quantitative comparison of image and laboratory spectra is done using spectral matching techniques, in particular the spectral angle (in the socalled spectral angle mapper algorithm ... Despite that there is very little to nothing reported on the overall performance of the various spectral matching techniques and on the means of assessing their performance. It is this niche that this paper intends to fill.”, and page 14 section 6 “Four spectral measures for matching known library spectra to unknown field or pixel spectra have been presented and their performance has been evaluated both on synthetic image data as well as on a AVIRIS data set” Van der Meer discloses the benefit of the method to provide several spectral matching techniques to determine similarity between spectrum feature and the technique to assess the performance of these technique. The assessment can be performed based on calculating a probability to indicate how confident the classification result is in classifying the pixel with its spectrum into a class. Given the technique to assess the performance of a classification result based on similarity with known reference data, one of ordinary skilled in the art may configure to improve a classification machine learning model based on the similarity comparison technique and the assessment of the similarity comparison, which help accurately determine the final class of a target. One of ordinary skilled in the art may further employ one or more similarity matching techniques and calculate the probability relate to each similarity matching technique to identify the most suitable technique that help obtain the most accurate class determination for a target. Therefore, the teaching of the classification model by Pabbaraju can be further improved in view of the method by Van der Meer.)
Regarding claim 3 depends on claim 1, thus the rejection of claim 1 is incorporated.
Pabbaraju the limitation “(e1) adding one to the number of votes for the classified class when the reliability degree is equal to or greater than a reliability degree threshold value, and invalidating a vote when the reliability degree is less than the reliability degree threshold value, for each of the N pieces of input data” (paragraph 21 “the system may determine the most voted class in the model output across all pixels”, and paragraph 46 “ A machine-learning algorithm 210 may generate a confidence level or factor for each output generated. For example, a confidence value that exceeds a predetermined high-confidence threshold may indicate that the machine-learning algorithm 210 is confident that the identified feature corresponds to the particular feature. A confidence value that is less than a low-confidence threshold may indicate that the machine-learning algorithm 210 has some uncertainty that the particular feature is present”. Pabbaraju discloses the confidence value exceeds a predetermined threshold may indicate that the machine learning algorithm is confident in its result. In the scheme of classification, it would mean an indication of confidence in the classified input pixel. Since Pabbaraju already implies that each classified class contain a number of votes from each classified pixel as the system may determine the most voted class across all pixels, one of ordinary skilled in the art may interpret that one count of vote is registered every time a pixel is classified into a class, wherein the classification process can be further enhanced with the confidence threshold comparison to accurately determine to register the vote or not. Pabbaraju also discloses the uncertainty in each vote of the pixel with its feature belong to a class. One of ordinary skilled in the art may configure to consider the vote with such uncertainty as invalid as it deems the classification as uncertain.)
Pabbaraju teaches the limitation “(e2) determining, as the class determination result, a class among the M classes, the class having the largest number of votes for the N pieces of input data.” (paragraph 31 “the majority-vote Superpixel loss may, for all pixels in the image, set the most voted class in the corresponding superpixels as the target.” Pabbaraju discloses the majority voting scheme for the pixels within the image, wherein the most voted class is determined to be the class of the image.)
Regarding claim 5 depends on claim 1, thus the rejection of claim 1 is incorporated.
Pabbaraju teaches the limitation “(e1) adding the reliability degree as a vote value for the classified class when the reliability degree is equal to or greater than a reliability degree threshold value, for each of the N pieces of input data” (paragraph 46 “A machine-learning algorithm 210 may generate a confidence level or factor for each output generated. For example, a confidence value that exceeds a predetermined high-confidence threshold may indicate that the machine-learning algorithm 210 is confident that the identified feature corresponds to the particular feature.” Pabbaraju discloses the confidence level or factor, which corresponds to the probability in Van der Meer’s disclosure based on the teaching combination as explained above, thus the probability can be used to determine the vote for the classified class, as the probability within the teaching by Van der Meer has already been used to reflect the confidence in classifying pixel into class based on its spectrum similarity. The probability can be compared to a threshold by Pabbaraju and if it is greater or equal, then it can be used to further strengthen the confidence determination in a vote.)
Claim 5 is further rejected under the same rationale of claim 3 because the claim recites similar limitations and processing steps.
Regarding claim 7 depends on claim 1, thus the rejection of claim 1 is incorporated.
Van der Meer teaches the limitation “The method according to claim 1, wherein the specific layer has a configuration in which a vector neuron arranged in a plane defined with two axes including a first axis and a second axis is arranged as a plurality of channels along a third axis being a direction different from the two axes, and the feature spectrum is any one of: (i) a first type of the feature spectrum obtained by arranging a plurality of element values of an output vector of a vector neuron at one plane position in the specific layer, over the plurality of channels along the third axis” (page 4 fig. 1 “Concept of imaging spectrometry (hyperspectral remote sensing). An area on the ground is imaged at high spectral resolution and spectral sampling such that after systems corrections a reflectance spectrum can be retrieved for each pixel that can be directly compared to field- or laboratory spectra of materials (e.g., minerals in geological applications) of interest” Van der Meer discloses image pixel obtained based on spectral sampling from an image of an area on the ground at high spectral resolution. The pixel is analogous to the vector neuron arranged in a plane defined with two axes within the claim, wherein the crosstrack and the along track within figure 1 is analogous to the first and second axis and the spectral band is analogous to the third axis in a different direction from the two axes. Figure 1 further demonstrate the arranging of a plurality of a plurality of element values at one plane position in the specific layer over the plurality of channels along the third axis within the claim, wherein the 3D layer structure image of the area on the ground with high spectral resolution corresponds to the specific layer, wherein the associated pixel spectrum of each pixel corresponds to element values of an output vector of a vector neuron, wherein each pixel contain the pixel spectrum is located at one plane position of the 2D plane defined by the crosstrack and along track within the 3D layer structure image of the area on the ground, and wherein the pixel are arranged as stacking over the plurality of spectral bands which corresponds to the plurality of channels along the third axis.)
Regarding claim 8,
Pabbaraju teaches the limitation “a memory configured to store the machine learning model” (paragraph 36 “The memory unit 208 may include volatile memory and non-volatile memory for storing instructions and data ... For example, the memory unit 208 may store a machine-learning model 210 or algorithm” Pabbaraju discloses the embodiment of the system comprises the memory unit with a memory to store the machine learning model.)
Pabbaraju teaches the limitation “a processor configured to execute a calculation using the machine learning model” (paragraph 62 “Processor 520 may be configured to read into memory 522 and execute computer-executable instructions residing in non-volatile storage 516 and embodying one or more ML algorithms and/or methodologies of one or more embodiments.” Pabbaraju discloses the processor configured to read into the memory to execute the ML algorithms, which involve the calculation within the ML algorithms.)
Claim 8 is further rejected under the same rationale of claim 1 because the claim recites similar limitations and processing steps to claim 1.
Regarding claim 9,
Pabbaraju teaches the limitation “A non-transitory computer-readable storage medium storing a computer program for causing a processor ...” (paragraph 64 “The program code may be distributed using a computer readable storage medium having computer readable program instructions thereon for causing a processor to carry out aspects of one or more embodiments. Computer readable storage media, which is inherently non-transitory, may include volatile and non-volatile, and removable and non-removable tangible media implemented in any method or technology for storage of information, such as computer-readable instructions, data structures, program modules, or other data.” Pabbaraju discloses The program code may be distributed using a computer readable storage medium having computer readable program instructions thereon for causing a processor to carry out aspects of one or more embodiments.)
Claim 9 is further rejected under the same rationale of claim 1 because the claim recites similar limitations and processing steps to claim 1.
Claims 4, 6 are rejected under 35 U.S.C. 103 as being unpatentable over Pabbaraju et.al (US 20230107917 A1), in view of Van der Meer et.al (NPL: The effectiveness of spectral similarity measures for the analysis of hyperspectral imagery) further in view of Cruz Mota et.al (US 20150326450 A1)
Regarding claim 4 depends on claim 3, thus the rejection of claim 3 is incorporated.
Pabbaraju/Van der Meer does not teach the limitation “The method according to claim 3, wherein (e2) includes determining that a class of the target object is unknown when the largest number of votes is less than a vote number threshold value”. However, Cruz Mota teaches this limitation (paragraph 99 “In some embodiments, an optimized vote count threshold may also be associated with the set of voters/classifiers. If the number of votes is at or above the threshold, the original classification may be validated. However, if the number of votes falls below the threshold, the original classification may be deemed a false positive” Cruz Mota discloses a method of voting optimization using classifier, wherein the method comprises detecting the false positive classification using an optimized vote count threshold. If the number of votes falls below the threshold, the original classification may be deemed a false positive. The false positive classification is analogous to the unknown class determination within the claim.)
Before the effective filing date, it would have been obvious to a person having ordinary skill in the art to combine the teaching of a method of classifying an image with pixels into one or more classes using a machine learning model classifier with machine learning algorithm configured as a neural network with layers by Pabbaraju, and the teaching of methods to classify image data, which include calculating a similarity measure between known library spectra to pixel spectra of an image and to assess the confidence of the similarity measure using probability by van der Meer, with the teaching of voting optimization with comparison to a vote count threshold to determine false positive classification by Cruz Mota. The motivation to do so is referred to in Cruz Mota’s disclosure (paragraph 51 “To reduce the number of false positives, a voting mechanism may be implemented within network 100 to validate a detected attack before the supervisory device is notified. ... According to various embodiments, such a voting mechanism may also be optimized to determine which voters should participate in a vote, thereby reducing network overhead as a result of a vote. To further reduce the change of a false positive, optimizations may also be made regarding how consensus is reached (e.g., by setting a threshold number of votes for a confirmation to occur, etc.)” Cruz Mota discloses the benefit of the voting optimization method in accounting for and reducing the number of false positive classification. The voting mechanism may be implemented to validate a classification result. Such a voting mechanism may also be optimized thereby reducing network overhead and a vote count threshold can be set to further optimize the voting scheme. Therefore, the voting scheme to determine the classified class of the pixel within the teaching by Pabbaraju may be further optimized based on the teaching by Cruz Mota for further accuracy in final classification result.)
Regarding claim 6 depends on claim 5, thus the rejection of claim 5 is incorporated.
Claim 6 is further rejected under the same rationale as claim 4 because the claim recites similar limitations and processing steps, thus the claim is similarly rejected according to claim 4.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DUY TU DIEP whose telephone number is (703)756-1738. The examiner can normally be reached M-F 8-4:30.
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, Alexey Shmatov can be reached at (571) 270-3428. 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.
/DUY T DIEP/Examiner, Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123
Prosecution Insights