DETAILED ACTION
Notice of Pre-AIA or AIA Status
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Notice to Applicants
2. This communication is in response to the application filled on 10/03/2024.
3. Claims 1-20 are pending.
4. Limitations appearing inside {} are intended to indicate the limitations not taught by said prior art(s)/combinations.
Information Disclosure Statement
5. The information disclosure statement (IDS) submitted on 10/03/2024 has been considered by the examiner.
Claim Rejections - 35 USC § 101
5. 35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
6. Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The analysis below follows the Subject Matter Eligibility Test (See flowchart in MPEP 2106).
Step 1: Is the claim to a process, machine, manufacture or composition of matter? YES.
Step2A, Prong 1: Does the claim recite an abstract idea, law of nature, or natural phenomenon? YES. Claim 1 recites “A computer-implemented method for automatically grading coins, the method comprising (1): training a machine learning model based at least in part on a plurality of first images respectively depicting a plurality of coins of a particular type, individual ones of the plurality of coins being manually assigned a respective coin classification (2); receiving a second image depicting a different coin of the particular type (3); performing an analysis of the second image based at least in part on the machine learning model (4); and automatically assigning a particular coin classification to the different coin based at least in part on the analysis of the second image. (5)” [emphasis added].
Limitation (1), (2), (3), (4), and (5) recite a mental process, directed toward the mental process grouping of abstract ideas (MPEP 2106.04(a)(2)). Alternatively, limitation (2) may be alternatively understood to recite a mathematical calculation, specifically, the “…training a machine learning model…”. The examiner specifically highlights July 2024 Subject Matter Eligibility Examples, pg. 4, claim 2 and pg. 5-10, which describe an analogous example, and the examiner has considered claim (1) as a whole as being directed toward a single abstract idea of a mental process analogous to pg. 8, par. 1 of the July 2024 Subject Matter Eligibility Examples. A person, based at least in part on a plurality of first images respectively depicting a plurality of coins of a particular type, individual ones of the plurality of coins being manually assigned a respective coin classification, can grade a coin. Furthermore, a person, receiving a second image depicting a different coin of the particular type, can perform an analysis of the second image based at least in part on the plurality of first images; and automatically assigning a particular coin classification to the different coin based at least in part on the analysis of the second image.
Step2A, Prong 2: Does the claim recite an additional elements that integrate the judicial exception into a practical application? NO. Limitations (1), (2), (3), and (4), recite additional elements that fail to integrate the judicial exception of a mental process into a practical application. Specifically, (1), recites “A computer-implemented…”, which constitutes mere instructions to apply the on a generic computer (MPEP 2106.05(f)). Limitation (2) recites “…training a machine learning model…” which constitutes a well understood, routine and conventional activity in the art of image processing (MPEP 2106.05(d)), and may also be understood to be directed toward mathematical calculations grouping of abstract ideas. Limitation (3) recites an insignificant data acquisition step (MPEP 2106.05(g)). Limitation (4), recites “…analysis… in part on the machine learning model”, which constitutes mere instructions to apply the mental process of grading coins to the particular field of machine learning, as well as a generic computer recitation (MPEP 2106.05(h), MPEP 2106.05(f)).
Step 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception? NO. The claim’s additional elements, as stated in Prong 2, do not amount to significantly more than the judicial exception. Limitations (1), (2), (3), (4), and (5) lack sufficient structure to amount to significantly more than the judicial exception as described in Step2A. Furthermore, training a machine learning model to classify images is well-understood and routine within the industry. Therefore, limitations (2), (4), and (5) use well-understood, routine, and conventional activities previously known to the industry, specified at a high level of generality, to accomplish the judicial exception.
Claim 2-20 recite additional limitations that fail to integrate the judicial exception into a practical application. Specifically, claim 2 recites “…determining a date of the different coin based at least in part on the second image; and wherein automatically assigning the particular coin classification to the different coin is further based at least in part on the date.”, which is directed toward the abstract idea of a mental process analogous to claim 1 (MPEP 2106.04(a)(2)). Claim 3 recites limitations analogous to claim 2, and is likewise directed toward a mental process (MPEP 2106.04(a)(2)). Claim 4 recites “determining a hue-saturation-lightness (HSL) value of the different coin based at least in part on the second image; and wherein automatically assigning the particular coin classification to the different coin is further based at least in part on the HSL value.”, which is directed toward the abstract idea of a mathematical calculation and/or mental process (MPEP 2106.04(a)(2)), since conversion into HSL can be performed by hand (e.g., “HSL, HSB and HSV color: differences and conversion” to Vargas). Claim 5 recites “…particular coin classification is a grade on the Sheldon coin grading scale.”, which constitutes selecting a particular data source and/or type (MPEP 2106.05(g). Claim 6 recites “articular coin classification indicates at least one of: toning, color, or eye appeal.”, which analogous to claim 5, constitutes selecting a particular data source and/or type (MPEP 2106.05(g). Claim 7 recites “the particular type is one or more of a penny, a quarter, or a dollar coin.”, which constitutes selecting a particular data source and/or type (MPEP 2106.05(g). Claim 8 recites “automatically identifying the particular type of the different coin by performing an initial analysis of the second image; automatically identifying a particular variety of the particular type of the different coin by performing an initial analysis of the second image; or automatically identifying a mint error on the different coin by performing an initial analysis of the second image.”, which are directed toward a mental process (MPEP 2106.04(a)(2)). Claim 9 recites “determining a correlation between the respective coin classification and a ratio of local maxima of harmonics in each of the plurality of first images.”, which is directed toward the abstract idea of a mathematical calculation (MPEP 2106.04(a)(2)). Claim 10 recites “comprising performing a Fourier transform on the cropped second image.”, which is directed toward the abstract idea of a mathematical calculation (MPEP 2106.04(a)(2)). Claim 11 recites “… identifying text from the different coin in the second image; and cropping the second image around the text.”, which is directed a mental process (MPEP 2106.04(a)(2)). Specifically, a person can cut out and/or crop an image of a coin around the text after identifying it. Claim 13 and 14 recite analogous limitations that constitute selecting a particular data source and/or type (MPEP 2106.05(g). Claim 15 recites “the machine learning model uses at least one of. a K-nearest neighbors algorithm, a support vector machine, or a neural network.”, which constitutes mere instructions to apply the judicial exception to a particular field (MPEP 2106.05(h)). Claims 16 and 19 recites additional limitations analogous to claim 1, with claim 19 additional recites “A system… at least one computing device…”, which constitutes a generic computer recitation (MPEP 2106.05(f)). Claim 17 recites “…determining that the different coin is incorrectly classified; and outputting an automatically determined classification based at least in part on the analysis, the automatically determined classification differing from the proposed classification.”, which are directed toward a mental process (MPEP 2106.04(a)(2)). Claim 18 and 19 recites analogous limitations, which are directed toward a mental process (MPEP 2106.04(a)(2)). Specifically, a person can output a confidence score for their analysis and determine if a coin is correctly or incorrectly classified based on said confidence.
At least for these reasons, claim 1 are ineligible under 35 U.S.C. 101.
Claim Rejections - 35 USC § 102
7. The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
8. Claims 1, 5-7, 13, 16, and 19 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by “Machine Assisted Grading of Rare Collectibles through the COINS framework” to Bassett (hereinafter Bassett).
9. Regarding Claim 1, Bassett discloses a computer-implemented method for automatically grading coins ([pg. 40, Machine Model Components, par 1, ln. 1 to pg. 41, par. 3, ln. 4] “The technical grading model has major three components: Input, Process and Output (Figure 3.4.2)… The Input Component of the technical grading model provides users with the flexibility of either digitally scanning their collectible item via a GUI interface or to submitting a captured digitized image in a 256 color, 96 pixels/inch GIF format…When using the system locally, that is the software components are installed on their local hard drive, a user can scan a collectible or capture an image with a digital camera and perform the entire grading process on their own computer. The results of the grading can be privately maintained in a locally stored database…The Process Component has two distinct approaches to processing the grading request of the user: summary grading and detailed grading. Summary grading involves the examination of the collectible as a whole unit rather than using a detailed feature set. The summarized approach, which was used in this research, is the quicker of the two but it is also the least reliable because it is subject to false readings in certain conditions. In this method the entire digitized image of the collectible item is examined by using Histogram Distance Detection comparison algorithms. The output of this system is a technical grade without the subjective qualities that human graders typically apply when grading. Technical grading refers to the strict adherence to certain grading rules without the inclusion of subjective qualities. A technical grade is determined solely by wear and defects that occur after striking.”), the method comprising:
training a machine learning model based at least in part on a plurality of first images respectively depicting a plurality of coins of a particular type ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8] “Train the system. The coin images to be tested by the machine-based system were matched against the average grades assigned and stored into the Java based system. As this system was designed for research purposes only and the number of trained samples was not large a elaborate SQL relational database was not utilized. Researchers doing future work in this area might want to extend the system to include an SQL database to accommodate a much larger training database with thousands of samples and to decrease the speed of evaluation… In order to ensure that the machine-based system was evaluating the images properly a series of five reliability tests was constructed. The machine-based system was tested with various combinations of images in the trained database to determine if reliability increased of decreased substantially as the number of trained images was significantly changed… a) Run 50 of the 105 pregraded training coin images through the system (and removing these 50 images from the training database leaving a net of 50 trained images) to determine the level of accuracy of the machine-system… e) Run the 20 coins that were evaluated by the third- party grading services against the machine-based system with a training database size of 105 that had the level of accuracy (as determined by tests a, b, c & d above). This test was done to see how closely the machine-based system was able to return grades to those assigned by the third-party grading services. Note: The images that were being tested were removed from the trained database so that exact image to images matches would not occur. Automated grading of coins. The machine-based system was setup to compare the extracted histogram values against a stored database of trained histogram value measurements. The image being evaluated was converted to Hue, Saturation and Brightness vectors (HSB vectors) and compared against the HSB vectors of the images stored in the training database. The template matching logic was programmed to detect the closest distance measurement between the image being evaluated and the images in the database. Each image in the database is represented using three primaries of the color space chosen. The most common color space used is RGB. Each color channel is quantized into m intervals. So the total number of discrete color combinations (called bins) n is equal to m3. For example, each channel is commonly quantized into 16 intervals. So we have 4096 bins in total. A color histogram H(M) is a vector (h1, h2,..., hn), where each element hj represents the number of pixels falling in bin j in image M. These histograms are the feature vectors (indexes) to be stored as the index of the image database. The 24-bit RGB color for each pixel will be counted in eight intensity bins in each dimension. Each color has eight bits (255 color intensity), which are sampled. This gives a three-dimensional color vector for each pixel, with three bits representing each color dimension. Then the RGB vectors can be transformed to HSB color vectors. Histograms will then be performed by calculating the L2 norm distance between the histograms. Similarity results from HSB and RGB histogram searches will be compared, and weighting the HSB components for better comparison are also compared. During image retrieval, a histogram is found for the image. A metric is used to measure the distance between the histograms of the image, which is being evaluated, and the stored images in the database. (If images are of different size, their histograms can be normalized.) Evaluated images with the smallest distance are retrieved from the database and noted as the closest match, or nearest neighbor. [41]… Number and characteristics of Samples… Twenty sample coin images of Lincoln Cents were assembled as the experimental control group. The control group of samples was selected as a result of sending the twenty physical coins out to third-party grading services to be professionally graded… See figure 3.4.3 for what an encapsulated Lincoln Cents looks like… Each coin sample used in this study is part of the Lincoln Cent family, which has been minted from 1909 to present. Only the obverse (front) of the Lincoln Cent was evaluated in all of the experiments of this study. The obverse was selected rather than the entire coin as there was a major design change that occurred in 1959 when the reverse of the coin was changed over from the “Wheat Ears” to the “Lincoln Memorial”. Figure 3.04.3 provides illustrative samples of the changes in the reverse designs. Each sample was unique in that each coin was represented by a different date & mintmark combination (see Appendix B1 for a list of the samples).”), individual ones of the plurality of coins being manually assigned a respective coin classification ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8] see specifically “…e) Run the 20 coins that were evaluated by the third- party grading services against the machine-based system with a training database size of 105…” and “…Each coin sample used in this study is part of the Lincoln Cent family…”);
receiving a second image depicting a different coin of the particular type ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8] see “…e) Run the 20 coins that were evaluated by the third- party grading services against the machine-based system with a training database size of 105 that had the level of accuracy (as determined by tests a, b, c & d above). This test was done to see how closely the machine-based system was able to return grades to those assigned by the third-party grading services…”, also see “…Note: The images that were being tested were removed from the trained database so that exact image to images matches would not occur…”, [pg. 77, Test A-E Results, par. 1, ln. 1 to pg. 76, par. 5, ln. 5] “Table 4.2.5 summarizes the results of reliability tests A – E and demonstrates the larger the trained database that the machine was working from the better the machine would perform in terms of an increasing explained variance and a decreasing error rate. The machine-based system returned a level of accuracy of at least 93.5% on all reliability tests, which were conducted when there were 50 or more trained images in the database. Table 4.2.5 shows the summarized results of all tests that were conducted; the detail of all tests is available in Appendix F… The larger the trained database in the machine-based system the better the system seemed to do at grading, as the level of explained variance increased from 91.86% to 95.60% as the number of trained images went from 55 to 105. Increases in the number of trained images also seemed to contribute to the increase in percentage of coins, which were exactly matched as this percentage increased from 40% to 51.43% as the number of trained images went from 55 to 105… Employ multiple grading algorithms. Presently just the histogram distance measure algorithm was used for this research with impressive accuracy results. Perhaps by adding several additional matching algorithms the results could be increased by a using an interpolation of the results. Standalone machine-based systems that used a single new algorithm would need to be constructed and tested first to measure reliability”);
performing an analysis of the second image based at least in part on the machine learning model ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8] see “a)… e)…” and “Automated grading of coins…”, [pg. 77, Test A-E Results, par. 1, ln. 1 to pg. 76, par. 5, ln. 5]); and
automatically assigning a particular coin classification to the different coin based at least in part on the analysis of the second image ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8], [pg. 77, Test A-E Results, par. 1, ln. 1 to pg. 76, par. 5, ln. 5]).
10. Regarding Claim 5, Bassett discloses the computer-implemented method of claim 1. Bassett further discloses wherein the particular coin classification is graded on the Sheldon coin grading scale ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8], [pg. 77, Test A-E Results, par. 1, ln. 1 to pg. 76, par. 5, ln. 5] see Table 4.2.5 Average Coin Grade, [pg. 68, par. 4, ln. 1-8] “The long-term goal of a machine-based grading system is to always return a consistent technical grade that has a high level of trusted reliability and ultimately becomes a baseline grade for human grading. The results of this research demonstrates that the machine-based system, which was developed for this study, has a reliability rating in excess of 95% when it is trained with a sufficient representation of coins in the 1 – 70 Sheldon scale. While these results are encouraging, still better results are most likely possible if the number of evaluation algorithms and the number of sample images in the trained database [15, 29] were to be increased.”).
11. Regarding Claim 6, Bassett discloses the computer-implemented method of claim 1. Bassett further discloses wherein the particular coin classification indicates at least one of: toning, color, or eye appeal ([pg. 36, par. 2, ln. 1 to par. 4, ln. 10] “According to the experts in the coin-collecting field some of the shortcomings of a fully automated system for grading collectibles include: Lack of a true picture of the eye-appeal [9] as the eye-appeal algorithm didn’t provide the entire picture of eye-appeal, which is arguably subjective from person to person. Toning can have a drastic affect on the image of a coin [40]. Toning could not be measured accurately by the system, as measuring toning in with an automated system is a complex task. Toning is a natural discoloration of a coin's surface by the atmosphere over a long period of time. Many collectors often consider toning to be attractive and desirable and they tend to prefer coins with natural toning. Toning is a subjective quality that is an important factor in determining the value of a coin. Toning colors of major concern are white, copper, nickel, and gold depending upon the metal composition of the coin. However these major colors can include: red, red brown, brown, white, full white, original color, dark color, light tone, pleasing tone, rainbow tone, unusual tone, dark fields and light fields.”, [pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8], [pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8], [pg. 77, Test A-E Results, par. 1, ln. 1 to pg. 76, par. 5, ln. 5] see Table 4.2.5 Average Coin Grade, [pg. 68, par. 4, ln. 1-8], [pg. 113, see RARE COIN GRADING SCALE, AU50, AU58, MS60, and MS65] “…AU 50 - Slight traces of wear on the highest points of the coin; may be dull with some evidence of luster under any toning… AU 58-This is oftentimes called a slider as it will appear to many observers to be uncirculated. Just the faintest wear on the highest points of the coin. Luster should be quite evident, although some toning can be apparent. Usually coins with poor eye appeal will not make the AU 58 grade… MS 60 - Mint State indicates a coin that has no wear and is uncirculated. It may have numerous bagmarks and/or be toned. MS 60 is the lowest quality of an uncirculated coin… MS 65 - This is the gem category. Coin should be fully struck with eye appeal. Either brilliant or toned but there should not be any unsightly marks or color that negates eye appeal. Any marks should be very minor in appearance. Prices spread out even further….”).
12. Regarding Claim 7, Bassett discloses the computer-implemented method of claim 1. Bassett further discloses wherein the particular type is one or more of: a penny, a quarter, or a dollar coin ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8]).
13. Regarding Claim 13, Bassett discloses the computer-implemented method of claim 1. Bassett further discloses wherein the plurality of first image and the second image depict a coin obverse ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8]).
14. Regarding Claim 16, the claim language is analogous to claim 1 with the exception of “…the second image being associated with a proposed classification…” and “…automatically determining whether the different coin is correctly classified with the proposed classification based at least in part on the analysis…”, wherein the remainder of the claim is analogous to claim 1. Bassett further discloses the second image being associated with a proposed classification ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8] see “…e)…” specifically the “expert-third party grading images” ), and automatically determining whether the different coin is correctly classified with the proposed classification based at least in part on the analysis ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8], [pg. 73, par. 1, ln. 1-10] “The twenty coins that came back from the third-party grading services were run against the training database of 105 images in the machine-based system. These images were then run through the machine-based system one at a time, with the results of 11 coin images (55.00%) coming back with exact grade matches while 9 images (45.00%) were not matched exactly. The average grade variance in this test was 3.11 points, which was calculated by taking the total of all the differences between the expected grade and the grade the machine returned and then dividing that by the number of incorrect grades. The correlation between the machine grading results and results of the expert third party grading services results were .995 with an explained variance, or level of accuracy, of .990 while the error rate was .010”, [pg. 74, Table 4.2.4]). Rejections analogous to claim 1 are further applicable to the remainder of claim 16.
15. Regarding Claim 17, Bassett further discloses determining that the different coin is incorrectly classified ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8], [pg. 73, par. 1, ln. 1-10], [pg. 74, Table 4.2.4]), and outputting an automatically determined classification based at least in part on the analysis, the automatically determined classification differing from the proposed classification ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8], [pg. 73, par. 1, ln. 1-10], [pg. 74, Table 4.2.4]).
16. Regarding Claim 19, the claim language is analogous to claim 1 with the exception of “A system for automatically grading coins, comprising: at least one computing device configured to at least:…” wherein the remainder of the claimed is analogous to claim 1. Bassett specifically discloses a system for automatically grading coins, comprising: at least one computing device configured to at least perform the method ([pg. 40, Machine Model Components, par 1, ln. 1 to pg. 41, par. 3, ln. 4]). Rejections analogous to claim 1 are further applicable to the remainder of claim 19.
Claim Rejections - 35 USC § 103
17. 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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
18. Claims 2-3, 8, and 15 are rejected under 35 U.S.C. 103 as being unpatentable over “Machine Assisted Grading of Rare Collectibles through the COINS framework” to Bassett, and further in view of U.S. Patent No. 9,378,604 to Rathjen et al. (hereinafter Rathjen).
19. Regarding Claim 2, Bassett discloses the computer-implemented method of claim 1. Bassett does not specifically disclose determining a date of the different coin based at least in part on the second image; and wherein automatically assigning the particular coin classification to the different coin is further based at least in part on the date.
However, Rathjen teaches determining a date of the different coin based at least in part on the second image; and wherein automatically assigning the particular coin classification to the different coin is further based at least in part on the date ([col. 10, ln. 64 to col. 11, ln. 12] “Coin Attribute 430 records may comprise entries indicating Attributes of coins and, where applicable, ranges of values for Attributes. Coin Attribute 430 records may comprise a table (or similar data structure) describing Attributes and value of Attributes which a coin may have, such as, for example, type, strike, surface preservation, luster, mint mark, legend, device, dentils, fields, edge, date, rim, motto, denomination, director's edge mark, engraver's mark, exergue, material, color, finish, damage (e.g. bent, gouged, spot, scratch, corrosion, hole, jewelry, stain, and mutilated), or any other property, subjective or not, that the user knows or expects would be of collectible value (e.g. some specific eye appeal). A Coin ID 420 record may be associated with a subset of Coin Attribute 430 records to record Attributes measured with respect to a coin and its Coin ID 420 record.”, [col. 12, ln. 62 to col. 13, ln. 33] “At block 1408, the pixels in the Search Area 415 in the Coin Image 425 are extracted or identified. At block 1410, OCR and/or OIR is performed on the pixels extracted or identified in block 1408. OCR may be performed without reference to a Benchmark Image 410, while OIR may be performed with reference to one or more Benchmark Images 510. Because these relatively computationally complex processes are performed only with respect to a subset of the Coin Image 425, it is possible to perform this process rapidly. To identify Search Area 415 within Coin Image 425 first requires determining the proper rotation of Coin Image 425 relative to Benchmark Image 410 (which process is discussed above). To determine a wear estimate, a number of lines or curves present in the pixels in the coin may be compared to a number of lines or curves present in the Benchmark Image 410 or an amount of contrast in the pixels relative to an amount of contrast in the Benchmark Image 410. To the extent that the OCR/OIR output may have a confidence value, the confidence value may be obtained at block 1412 and OCR/OIR output below a confidence threshold may be rejected or flagged (and/or the confidence value may be stored in association with the Coin ID 420 record) while, at block 1414, the OCR/OIR output above the confidence threshold may be stored, for example, in Coin Attribute 530 records associated with the Coin ID 420 record. For example, the OCR/OIR output may comprise a number, such as a year, or a match with a mint-mark or a symbol. At block 1416, the color, color uniformity, and/or contrast of all or a part of Coin Image 425 may be determined. At block 1418, a numismatic value for the coin may be determined based on the Coin Attributes 430 determined in the preceding Blocks. The value may be obtained by looking up a price based on the Coin Attributes 430, such as in Price-Attribute Map 445 records. Price-Attribute Map 445 records may be clustered according to Human Coin Assessment 450 records associated with a Coin ID 420. Price-Attribute Map 445 records may comprise, for example, a “face” value of a coin, such as a stated currency amount, with adjustments (up or down) based on various of the Coin Attributes 430.”). One of ordinary skill in the art, before the effective filing date of the claimed invention, would specifically recognize Bassett and Rathjen as within the same field of image processing for numismatics, and as analogous to the claimed invention. The motivation to combine would have been obvious to one of ordinary skill in the art, in that a coin’s date is particularly relevant for numismatics, and in some cases can be directly tied to the value of the coin. One of ordinary skill in the art, before the effective filing date of the claimed invention, would have combined the method of Bassett with the coin date determination of Rathjen through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. Specifically, one of ordinary skill in the art would have modified the method of Bassett to further include determining a date on a coin and automatically assigning the particular coin classification to the different coin is further based at least in part on the date as taught in Rathjen (e.g., older coin may be classified as higher grade then new mint coin, etc.)
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the method of Bassett with the coin date determination of Rathjen to obtain the invention as specified in claim 2.
20. Regarding Claim 3, Bassett discloses the computer-implemented method of claim 1. Bassett does not specifically disclose determining a level of wear of the different coin based at least in part on the second image; and wherein automatically assigning the particular coin classification to the different coin is further based at least in part on the level of wear.
However, Rathjen teaches determining a level of wear of the different coin based at least in part on the second image; and wherein automatically assigning the particular coin classification to the different coin is further based at least in part on the level of wear ([col. 10, ln. 64 to col. 11, ln. 12], [col. 12, ln. 62 to col. 13, ln. 33]). The motivation remains analogous to claim 2. One of ordinary skill in the art, before the effective filing date of the claimed invention, would have combined the method of Bassett with the coin wear determination of Rathjen through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. Specifically, one of ordinary skill in the art would have modified the method of Bassett to further include determining a wear on a coin and automatically assigning the particular coin classification to the different coin is further based at least in part on the wear as taught in Rathjen (e.g., coin with more wear graded lower than coin with no wear, etc.)
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the method of Bassett with the coin wear determination of Rathjen to obtain the invention as specified in claim 3.
21. Regarding Claim 8, Bassett discloses the computer-implemented method of claim 1. Bassett does not specifically disclose automatically identifying the particular type of the different coin by performing an initial analysis of the second image; automatically identifying a particular variety of the particular type of the different coin by performing an initial analysis of the second image; or automatically identifying a mint error on the different coin by performing an initial analysis of the second image. Specifically, Bassett uses only one type of coin, and therefore, does not need to identify a type of coin by performing analysis.
However, Rathjen specifically teaches at least one of: automatically identifying the particular type of the different coin by performing an initial analysis of the second image; automatically identifying a particular variety of the particular type of the different coin by performing an initial analysis of the second image; or automatically identifying a mint error on the different coin by performing an initial analysis of the second image ([col. 10, ln. 31-36] “Coin Type 407 may comprise a type of coin, such as a penny, nickel, quarter, or the like, associated with a set of Recognition Criteria 405. Coin Type 407 may comprise narrower types within a broader type, such as a narrower “Wheat Cent” Coin Type 407, within the broader “US Penny” Coin Type 407.”, [col. 12, ln. 6-42] “At block 1340, the Recognition Criteria 405 of block 1335 are applied to Coin Image 425, for example, to match Coin Image 425 to a Benchmark Image 410 record or to match Coin Image 425 to template-independent criteria (not already obtained at block 1315). Block 1340 may proceed by taking the central portion of Coin Image 425 from Block 1330, and match it to Recognition Criteria 405, such as, for example, by overlaying Coin Image 425 pixels on Recognition Criteria 405 pixels and subtracting (and potentially squaring) the values; if the result is not zero or not close to zero, then no match is determined. Alternative matching algorithms may be followed, such as algorithms which involve pixel addition or algorithms which involve determining a hash of Coin Image 425 across a matrix of pixel values which matrix has an orientation and comparing the hash value to a corresponding hash value of Recognition Criteria 405 (or of Benchmark Image 410). If no match is determined at block 1340, and if at block 1345 a full set of rotations has not yet been completed (in which case, an error or a “no result” message may be returned and the coin ejected at an associated Bin 170), then at block 1350 the central portion of Coin Image 425 from block 1330 may be rotated (or the pixel values in the matrix are rotated), such as, for example, by one degree. The process may then return to block 1340 to determine if a match is found. If at block 1345 a full set of rotation has been completed and “no result” is returned, then the lack of a result may be identified as a Coin Type 407, such as an “unknown” Coin Type 407. At block 1355, if a match is found at block 1340 or proceeding from block 1320, a Coin Type 407 may be obtained, which Coin Type 407 is associated with Recognition Criteria 405 and/or Benchmark Image 410 records. At block 1360, Coin Type 407, Benchmark Image 410, and/or Recognition Criteria 405 records may be associated with Coin ID 420 and/or loaded in a reduced set of Recognition Criteria 405 to be used in the second pass of Recognizer 500, blocks 528 to 554.”). The motivation to combine would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, in that by identifying the type of coin, you could expand the usage to multiple types of coins automatically. One of ordinary skill in the art, before the effective filing date of the claimed invention, would have combined the method of Bassett with the coin type determination of Rathjen through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. Specifically, one of ordinary skill in the art would have modified the method of Bassett to further include determining a type of coin (e.g., cent vs. dollar), a particular variety (e.g., silver dollar vs. gold dollar), and a mint error (e.g., fault mints, such as multiple stamps and/or of center marks) as taught in Rathjen.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the method of Bassett with the coin type determination of Rathjen to obtain the invention as specified in claim 8.
22. Regarding Claim 15, Bassett discloses the computer-implemented method of claim 1. Bassett does not specifically disclose wherein the machine learning model uses at least one of a {K-}nearest neighbors algorithm, {a support vector machine, or a neural network} ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8] see “…Evaluated images with the smallest distance are retrieved from the database and noted as the closest match, or nearest neighbor…”). Bassett does not specifically disclose wherein this is K-nearest neighbors’ algorithm.
However, Rathjen specifically teaches wherein the machine learning algorithm uses a neural network ([col. 16, ln. 63 to col. 17, ln. 16] “At block 1005, Machine Learning 1000 routine receives a Coin Image 425. At block 1010, Machine Learning 1000 routine executes a machine learning process on the received Coin Image 425. The machine learning process may be a supervised process involving statistical classification or an unsupervised learning process involving an artificial neural network, association rule learning, hierarchical clustering, cluster analysis, and/or outlier detection. The machine learning process may involve reinforcement learning. At block 1015, Machine Learning 1000 routine outputs Coin Attributes 430 for the coin according to the learning process of block 1010. At block 1020, Machine Learning 1000 routine may receive a human assessment and/or human assigned coin attributes for the coin, such as Human Coin Assessment 450. At block 1025, Machine Learning 1000 routine may provide feedback to the machine learning process of block 1010, in the form of Human Coin Assessment 450 of block 1020. The feedback may reinforce determinations by Machine Learning 1000 routine which were correlated with Human Coin Assessment 450.”). The motivation to combine would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, in that a neural network is substitutable for the nearest neighbor analysis of Bassett and offers an automatic means to classify coins. One of ordinary skill in the art, before the effective filing date of the claimed invention, would have combined the method of Bassett with the neural network of Rathjen through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. Specifically, one of ordinary skill in the art would have modified the method of Bassett to operate using a neural network (e.g., by using a known pre-trained model such as VGG-16 or ResNet) as taught in Rathjen.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the method of Bassett with the neural network of Rathjen to obtain the invention as specified in claim 15.
23. Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over “Machine Assisted Grading of Rare Collectibles through the COINS framework” to Bassett, and further in view of “HSL, HSB and HSV color: differences and conversion” to Vargas (hereinafter Vargas).
24. Regarding Claim 4, Bassett discloses the computer-implemented method of claim 1. Bassett further discloses determining a hue-saturation-{lightness} (HS{L}) value of the different coin based at least in part on the second image; and wherein automatically assigning the particular coin classification to the different coin is further based at least in part on the HS{L} value ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8]). Bassett fails to specifically teach hue-saturation-lightness values.
However, Vargas teaches HSL values ([pg. 1, par. 1, ln. 1-5] “If you are here you've probably heard of HSL, HSB or HSV color models. They are nothing more than transformations of RGB values. Simple equations are used to convert from RGB to HSL and HSV color. HSB is exactly the same as HSV, but HSL is slightly different as explained below. Because they are both based on RGB, they are also meant to be used to represent color in digital devices. This article explains the basic definitions involved (no previous knowledge on color theory is assumed) and shares the equations to convert between them.”, [pg. 1, par. 2, ln. 1-3] “Why did they bother creating color models that represent the same colors RGB already could? Although RGB is perfect for machines, it is not very human-friendly. HSL and HSV color models were created as a more convenient way for us to specify colors in software.”, [pg. 1, par. 5, ln. 1-5] “HSL is slightly different. Hue takes exactly the same numerical value as in HSB/HSV. However, S, which also stands for Saturation, is defined differently and requires conversion. L stands for Lightness, is not the same as Brightness/Value. Brightness is perceived as the "amount of light" which can be any color while Lightness is best understood as the amount of white. Saturation is different because in both models is scaled to fit the definition of brightness/lightness.”, [pg. 3, Converting Between HSB/HSV and HSL, see conversion equations]). One of ordinary skill in the art, before the effective filing date of the claimed invention, would recognize the HSB vector and color histogram analysis of Bassett to be directly analogous to an HSL value of the different coin. Specifically, one of ordinary skill in the art, before the effective filing date of the claimed invention, would recognize HSL to be substitutable for HSB, specifically in view of them both being spherical representations of the RGB color space based on hue and saturation, and given that you can easily convert between the two-color spaces with minimal effort. Specifically, Vargas further teaches such a conversion and explains the slight differences between HSL and HSB in such a manner that it would have been apparent to one of ordinary skill in the art that you can substitute the HSB values of Bassett for HSL values as taught in Vargas ([pg. 1, par. 1, ln. 1-5], [pg. 1, par. 2, ln. 1-3], [pg. 1, par. 5, ln. 1-5], [pg. 3, Converting Between HSB/HSV and HSL, see conversion equations]). One of ordinary skill in the art, before the effective filing date of the claimed invention, would have substituted the HSB value of the method of Bassett with the HSL value as taught in Vargas through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. The motivation to combine would have been obvious to one of ordinary skill in the art, and is disclosed in Vargas, in that HSL could allow for an easier understanding for a user ([pg. 1, par. 2, ln. 1-3]).
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the method of Bassett with the HSL value as taught in Vargas to obtain the invention as specified in claim 4.
25. Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over “Machine Assisted Grading of Rare Collectibles through the COINS framework” to Bassett, and further in view of “Local Image Patterns for Counterfeit Coin Detection and Automatic Coin Grading” to Gakhar (hereinafter Gakhar).
26. Regarding Claim 9, Bassett discloses the computer-implemented method of claim 1. Bassett does not specifically disclose determining a correlation between the respective coin classification and a ratio of local maxima of harmonics in each of the plurality of first images.
However, Gakhar specifically teaches determining a correlation between the respective coin classification and a ratio of local maxima of harmonics in each of the plurality of first images ([pg. 18, 4.1 Feature Extraction Using SIFT, par. 1, ln. 1 to pg. 20, par. 1, ln. 4] “SIFT image features are free from many complexities found in other methods such as object rotation and scaling, and they are resistant to any kind of noise in the image. The SIFT method converts the whole image into "Group of local feature vectors" [27]. Every feature obtained is a scale and rotation invariant… In this stage, we obtain the location and scale of the same object with different views by using the "scale-space" function, and as per assumptions, it is based on the Gaussian function. Out of several techniques available to find stable key Points in scale space. The difference of Gaussians is one of the methods to find the scale-space extrema, 𝐷(𝑥,𝑦,𝜎) by calculating the difference between the two images, where we have one image with scale k times the other. 𝐷(𝑥, 𝑦,𝜎) is then given by: 𝐷(𝑥,𝑦,𝜎)= L(𝑥,𝑦,k𝜎)- L(𝑥,𝑦,𝜎). We detect local maxima and minima of 𝐷(𝑥,𝑦,𝜎) and compare it with its 8 neighbors that are at the same scale as well as nine neighbors from above and below of one scale. The point is taken as an extremum If this value is the minimum or maximum of all these points. At this stage, we remove key points that are poorly localized or have low contrast from the list of key points extracted. We find the Laplacian value for every key point in stage 1. We take the location of extremum z, and the point is left out if it's below the threshold value when taking function value at z. In this way, we remove extremes with low contrast from the set of points. We consider a sizable principal curvature across the edge to remove poorly localized points, but there is also a small curvature in the perpendicular direction in the difference of Gaussian function. We reject the key point if the difference is lower than the ratio of the largest to the smallest eigenvector, from the 2x2 Hessian matrix at the location and scale of the key point. This stage takes into account the local image properties, the key points are assigned a consistent orientation, and then a key point can be represented relative to it, making it invariant to rotation. Orientation can be assigned by selecting the Gaussian smoothed image L from above by using the key points scale, compute gradient magnitude m, and compute orientation θ. Gradient orientations of the sample points form an orientation histogram. We find the highest peak in the histogram. This peak and any other peak within 80% height of this peak create a key point with that orientation.”). One of ordinary skill in the art, before the effective filing date of the claimed invention, would specifically recognize Bassett and Gakhar as within the same field of image processing for numismatics, and as analogous to the claimed invention. The motivation to combine would have been obvious to one of ordinary skill in the art, and is taught in Gakhar, wherein you obtain features that are scale invariant and rotation invariant ([pg. 18, 4.1 Feature Extraction Using SIFT, par. 1, ln. 1 to pg. 20, par. 1, ln. 4]), which means that images don’t need to be aligned perfectly to determine similarity. One of ordinary skill in the art, before the effective filing date of the claimed invention, would have combined the method of Bassett with the ratio of local maxima of harmonics as taught in Gakahr, through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. Specifically, one of ordinary skill in the art would have modified the method of Basset to further determining a correlation between the respective coin classification and a ratio of local maxima of harmonics in the first images as taught in Gakhar.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the method of Bassett with the ratio of local maxima of harmonics as taught in Gakahr to obtain the invention as specified in claim 9.
27. Claim 10-11, and 14 are rejected under 35 U.S.C. 103 as being unpatentable over “Machine Assisted Grading of Rare Collectibles through the COINS framework” to Bassett, and further in view of “Automatic Coin Classification and Identification” to Huber-Mork et al. (hereinafter Huber).
28. Regarding Claim 10, a Bassett discloses the computer-implemented method of claim 1. Bassett does not specifically disclose wherein performing the analysis of the second image further comprises: applying a transform to straighten a feature of the different coin; and cropping the second image around the feature of the different coin.
However, Huber specifically teaches applying a transform to straighten a feature of the different coin ([pg. 133, Fig. 4], [pg. 133, par. 2, ln. 1-13] “Rotational invariance for a coin edge image involves cross-correlation with reference edge images. The edge image is mapped from Cartesian to polar coordinates, see Fig.4. The result of cross-correlation between the coin image to be classified and a set of reference images is used to derive class hypotheses. In detail, for both sides of a coin under investigation rotational invariant processing and hypothesis generation proceeds as follows: 1. Estimation of coin diameter from coin detection. 2. Selection of a set of reference images depending on thickness and diameter measure (if available). Each reference image is associated with a coin class. 3. Cross-correlation of the coin side edge image under investigation with all reference coin edge images in the selected reference set, resulting in a cross-correlation value and associated rotation angle estimation for each reference class. 4. Ranking of the reference set by the maximum correlation value and generation of a set of hypotheses for the highest-ranking classes.”, [pg. 133, par. 3, ln. 1 to pg. 134, par. 1, ln. 3] “To obtain reliable estimates for cross-correlation and rotation angle the polar image is split into n bands along the radius coordinate, corresponding to concentric rings in Cartesian coordinates. The peak of the correlation value
K
i
for band i is determined for each band and the position of the peak is taken as an estimate for the rotation angle in band i. The sample mean angle direction
α
-
is estimated via (Fisher, 1995):
α
-
=
a
r
c
t
a
n
S
C
i
f
S
≥
0
a
n
d
C
>
0
a
r
c
t
a
n
S
c
+
π
i
f
C
<
0
a
r
c
t
a
n
S
c
+
2
π
i
f
S
<
0
a
n
d
C
<
0
with
C
=
∑
i
=
1
n
δ
i
c
o
s
α
i
,
S
=
∑
i
=
1
n
δ
i
s
i
n
α
i
.
If band i contains a significant number of edge pixels in reference coin and coin under investigation
δ
i
=
1
, otherwise
δ
i
=
0
. A cross-correlation estimate K for the coin under investigation is calculated using
K
=
1
/
n
'
∑
i
=
1
n
δ
i
K
i
. The number of bands
n
'
≤
n
used in cross-correlation and angle estimation varies between images and is simply obtained by
n
'
=
∑
i
=
1
n
δ
i
.”); and cropping the second image around the feature of the different coin ([pg. 133, Fig. 4] see specifically “Split into bands with overlap” and “Fourier transform”, see also [pg. 133, par. 3, ln. 1 to pg. 134, par. 1, ln. 3]). One of ordinary skill in the art, before the effective filing date of the claimed invention, would specifically recognize Bassett and Huber as within the same field of image processing for numismatics, and as analogous to the claimed invention. The motivation to combine is disclosed in Huber, wherein taking into account rotation is necessary for recognition and matching ([pg. 132, 3.2 Invariant preprocessing for 2D images, par. 1, ln. 1 to par. 2, ln. 4] “Apart from illumination dependency the appearance of a coin varies considerably with respect to its grey values depending on dirt and abrasion. These variations frequently are inhomogeneous. This suggests, even if illumination influence could be neglected, that for recognition purposes grey values by themselves will not give appropriate results. On the other hand, edge information remains more or less stable or at least degrades gracefully. Therefore, we based the feature extraction for coin recognition on edges… For reliable matching of coins invariance with respect to rotation has to be taken into account. Invariance with respect to translation is already discussed and taken into account by an approach involving segmentation in Sec. 3.1. Scale variance is accounted for either by using a calibrated acquisition device or normalization of the segmented image.”). One of ordinary skill in the art, before the effective filing date of the claimed invention, would have combined the method of Bassett with the transform and cropping as taught in Huber, through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. Specifically, one of ordinary skill in the art would have modified the method of Bassett by applying a transform to straighten a feature of the different coin and cropping the second image around the feature of the different coin as taught in Huber.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the method of Bassett with the transform and cropping as taught in Huber to obtain the invention as specified in claim 10.
29. Regarding Claim 11, a combination of Bassett and Huber teaches the computer-implemented method of claim 10. Bassett does not specifically disclose performing a Fourier transform on the cropped second image.
However, Huber specifically discloses performing a Fourier transform on the cropped second image ([pg. 133, Fig. 4], [pg. 133, par. 2, ln. 1-13], [pg. 133, par. 3, ln. 1 to pg. 134, par. 1, ln. 3]). The motivation to combine remains analogous to claim 10. One of ordinary skill in the art, before the effective filing date of the claimed invention, would have combined the method of Bassett with the transform, cropping, and Fourier transform as taught in Huber, through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. Specifically, one of ordinary skill in the art would have modified the method of Bassett to further apply a Fourier transform on the cropped second image as taught in Huber.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the method of Bassett with the transform, cropping, and Fourier transform as taught in Huber to obtain the invention as specified in claim 11.
30. Regarding Claim 14, Bassett does not specifically disclose wherein the plurality of first images and the second image depict a coin reverse.
However, Huber specifically teaches wherein the plurality of first images and the second image depict a coin reverse ([pg. 128, par. 3, ln. 1-10] “We will compare two approaches for classification of coins, a method based on matching edge features in polar coordinates representation (Nölle et al., 2003) and a method for matching based on an Eigenspace representation (Huber et al., 2005)… Classification of modern coins makes additional use of geometric measurements and information extracted from obverse and reverse side of the coins. Incorporation of geometrical measurements and fusion of coin sides is realized by preselection and Bayesian fusion…”, [pg. 128, par. 5, ln. 1-3] “Results are presented for all considered data sets and methods. The data set for classification of coins consisted of approximately12 000 coins with images of reverse and obverse sides. The data set contained 932 different coin classes…”). The motivation to combine would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, in that the reverse side contains important numismatic information that is directly relevant to classification, and thus not including a coin reverse could result in faulty or ineffective classification. One of ordinary skill in the art, before the effective filing date of the claimed invention, would have combined the method of Bassett with the coin reverse images as taught in Huber, through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. Specifically, one of ordinary skill in the art would have modified the method of Bassett to further operate with coin reverse images as taught in Huber.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the method of Bassett with the coin reverse images as taught in Huber to obtain the invention as specified in claim 14.
31. Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over “Machine Assisted Grading of Rare Collectibles through the COINS framework” to Bassett, and further in view of “Topology-Based Character Recognition Method for Coin Date Detection” to Pan et al. (hereinafter Pan).
32. Regarding Claim 12, Bassett discloses the computer-implemented method of claim 1. Bassett does not specifically disclose wherein performing the analysis of the second image further comprises: identifying text from the different coin in the second image; and cropping the second image around the text.
However, Pan specifically discloses wherein performing the analysis of the second image further comprises: identifying text from the different coin in the second image; and cropping the second image around the text ([pg. 1834, Fig. 4], [pg. 1835, Fig. 5 and 6], [pg. 1834, III. Coin Date Extraction, par 1, ln. 1 to 1835 col. 1, par. 2, ln. 13] “Date extraction is carried out by searching the possible date zone in the coin photo. Even through a date could appear at any position of each side of the coin, in most cases its position is relatively limited especially when we know some of a priori knowledge such as the country of the given coin. We could also consider that a pre-selection as being done using [2] for example. Therefore, a list of candidate zones is proposed in order to reduce the searching time over the entire image. According to our observations on more than 3000 modern American coins, the bottom of the obverse and the middle of the reverse have the most change to contain the date. As sometimes the date on the bottom runs in a circular way inside the coin border, a polar coordinates transform is used to put the circular date in a horizonal way. Thus, if we have no idea which side of the coins is in the image, the candidate zones are defined at the bottom part of
z
1
c
and
z
1
p
and in the middle part by
z
2
. Let us denote by
Z
=
{
z
1
c
,
z
1
p
,
z
2
}
the set of candidate zones (see Fig. 4). Let us define W the set of all possible sliding windows in a candidate zone
z
i
∈
Z
and
l
m
i
n
,
l
m
a
x
,
h
m
i
n
,
h
m
a
x
,
r
m
i
n
, and
r
m
a
x
the six extrema of width, height and aspect ratio for a possible date zone
w
d
a
t
e
∈
W
. For each sliding window
w
i
, we compute a normalized histogram
H
(
w
i
)
obtained by scanning vertically then horizontally the gradient map of the image inside the window… The second line of Fig. 5 illustrates some typical imperfect extraction: (d) non-horizontal extraction; (e) partial extraction lacking a small part of numbers; (f) partial extraction lacking a complete number… The extracted date is then cropped into four individual numbers to recognize. The crop is simply proceeded by searching the “valleys” of the histogram
H
(
w
d
a
t
e
)
which are closest to the supposed separations… For individual cropped number images, we could have imperfect but workable samples including slightly skewed numbers…”). One of ordinary skill in the art, before the effective filing date of the claimed invention, would specifically recognize Bassett and Pan as within the same field of image processing for numismatics, and as analogous to the claimed invention. The motivation to combine would have been obvious to one of ordinary skill in the art, and is disclosed in Pan, wherein the date information is particularly important to identify coin ([pg. 1833, Abstract, par. 1, ln. 11] “For recognizing coins, the graved release date is important information to identify precisely its monetary type. However, reading characters in coins meets much more obstacles than traditional character recognition tasks in other fields, such as reading scanned documents or license plates. To address this challenging issue in a numismatic context, we propose a training-free approach dedicated to detection and recognition of the release date of the coin. In the first step, the date zone is detected by comparing histogram features; in the second step, a topology-based algorithm is introduced to recognize coin numbers with various font types represented by binary gradient map.”). One of ordinary skill in the art, before the effective filing date of the claimed invention, would have combined the method of Bassett with the text identification and cropping as taught in Pan, through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. Specifically, one of ordinary skill in the art would have modified the method of Bassett to further identifying text from the different coin in the second image; and crop the second image around the text as taught in Pan.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the method of Bassett with the text identification and cropping as taught in Pan to obtain the invention as specified in claim 12.
33. Claim 18 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over “Machine Assisted Grading of Rare Collectibles through the COINS framework” to Bassett, and further in view of “Deep ancient Roman Republican coin classification via feature fusion and attention” to Anwar et al. (hereinafter Anwar).
34. Regarding Claim 18, Bassett discloses the computer-implemented method of claim 16. Bassett further discloses automatically determining whether the different coin is correctly classified with the proposed classification is based at least in part on {a confidence level associated with} the analysis ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8], [pg. 73, par. 1, ln. 1-10] [pg. 74, Table 4.2.4]). Bassett does not specifically disclose a confidence level associated with the analysis.
However, Anwar specifically discloses a confidence level associated with the analysis ([pg. 8, Fig. 7] “The correctly classified images are represented with green circles while the wrongly classified ones are in red circles. In the first row, the confidence of the NasNet [44] is always low, although the model can classify correctly. The second shows the confidence of the VGG [45] , which is consistently high even for wrongly classified classes. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)” see confidence scores above coins, [pg. 8, col. 1, 5.3. Qualitative comparison, par. 1, ln. 1 to col. 2, par. 1, ln. 9] “In Fig. 7 , we show the correct and incorrect classification results on the randomly selected images from the original test dataset. The results are only furnished for the CNN algorithms’ i.e. VGG [45] , NasNet [44] and our CoinNet. In the top row of Fig. 7 , our method, and NasNet [44] both can classify the input images correctly; hence marked with different shades of green circles and a confidence score at the top of each image. It can be observed that the confidence level of NasNet [44] is always lower even the prediction is correct compared to our CoinNet method. Likewise, we present the misclassification of the coin types by our method and VGG [45] in the bottom row of Fig. 7 marked with red circles and again having the confidence score at the top of the image. In this case, VGG [45] is always more confident i.e., having a high score than our network. This sums up that our model is more confident about correct predictions and vice versa.”). One of ordinary skill in the art, before the effective filing date of the claimed invention, would specifically recognize Bassett and Bassett as within the same field of image processing for numismatics, and as analogous to the claimed invention. The motivation to combine would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, in that a confidence score can assist in understanding and allow for comparison with other models and/or human scoring. One of ordinary skill in the art, before the effective filing date of the claimed invention, would have combined the method of Bassett with the confidence level with the analysis of Anwar, through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. Specifically, one of ordinary skill in the art would have modified the method of Bassett to further output a confidence level associated with the analysis as taught in Anwar.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the method of Bassett with the confidence level with the analysis of Anwar to obtain the invention as specified in claim 18.
35. Regarding Claim 20, Bassett discloses the system of claim 20. Bassett further discloses determining whether a proposed coin classification is incorrect based at least in part on the particular coin classification and {a confidence level associated with} the analysis ([pg. 42, par. 3, ln. 1 to pg. 44, par. 2, ln. 8], [pg. 73, par. 1, ln. 1-10] [pg. 74, Table 4.2.4]). Specifically, the examiner notes that Bassett also determines if a classification is incorrect ([pg. 73, par. 1, ln. 1-10]). Bassett does not specifically disclose a confidence level associated with the analysis.
However, Anwar specifically discloses a confidence level associated with the analysis ([pg. 8, Fig. 7], [pg. 8, col. 1, 5.3. Qualitative comparison, par. 1, ln. 1 to col. 2, par. 1, ln. 9]). The motivation to combine remains analogous to claim 18. Specifically, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, that the confidence level could be used to determine whether a proposed coin classification is incorrect (e.g., low confidence for classification would indicate possible misclassification, see bottom examples of misclassifications in [pg. 8, Fig. 7] of Anwar). One of ordinary skill in the art, before the effective filing date of the claimed invention, would have combined the method of Bassett with the confidence level with the analysis of Anwar, through known means, with no change to their respective function, and the combination would have yielded nothing more than predicable results. Specifically, one of ordinary skill in the art would have modified the method of Bassett to further output a confidence level associated with the analysis as taught in Anwar, and determine if a coin classification is incorrect based on the confidence level (e.g., if confidence is low, output is likely incorrect, etc.).
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine the system of Bassett with the confidence level with the analysis of Anwar to obtain the invention as specified in claim 20.
Conclusion
36. The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure. See PTO-892.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PAULO ANDRES GARCIA whose telephone number is (703)756-5493. The examiner can normally be reached Mon-Fri, 8-4:30PM ET.
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, Chan Park can be reached on (571)272-7409. 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.
/PAULO ANDRES GARCIA/Examiner, Art Unit 2669
/SUMATI LEFKOWITZ/Supervisory Patent Examiner, Art Unit 2672
Prosecution Insights