DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
A Preliminary Amendment was made 02/16/2024 to amend claims 3-8, 11-14, 18, 21 from pending claims 1-21.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 02/106/2024, 06/24/2024, 01/07/2025, 01/14/2026 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are considered by examiner.
Claim Objections
Claim 21 is objected to because of the following informalities: Claim 21 is objected to for the equation resolution being too low in resolution leading to an illegible submission. The equation is presumed to be equation 1 as described in the specification (¶ [0061]). However, the applicant is required to confirm the equation claimed and submit the claim with legible characters representing the function.. Appropriate correction is required.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-5, 8-16 are rejected under 35 U.S.C. 103 as being unpatentable over Cohen et al (US 2020/0234417, disclosed in IDS dated 06/24/2024) in view of Narasimhan et al (US 2022/0067911, disclosed in IDS dated 02/16/2024).
Regarding Claim 1, Cohen et al teach a method for classifying a defect in a manufactured article performed by a defect classification system (method of examining a specimen via an image for a defect; Fig 2 and ¶ [0061]), comprising:
receiving inspection data generated by an inspection tool inspecting the manufactured article (an inspection image of a die of a specimen can be captured 202 by an examination tool 120; Fig 2 and ¶ [0062]);
generating a defect map with the inspection data, the defect map including an arrangement of potential defect indicators (generating a defect map indicative of defect candidate distribution on the inspection image using reference images 202; Fig 2 and ¶ [0062]-[0063], [0066]);
modifying a subset of the potential defect indicators based on at least one spatial attribute of the arrangement to generate a modified defect map (a plurality of defect candidates can be selected 204 based on predefined attributes, such as a specific pattern (spatial attribute) and a modified inspection image patch can be generated 206; Fig 2 and ¶ [0067]-[0068]); and
classifying, with the machine learning model, the defect in the modified defect map, (the attributes of the defects detected in the defect map based on attribute space may be grouped in clusters based on machine learning techniques and a defect re-detection can be performed to identify and cluster groups; ¶ [0102]-[0103]).
Cohen et al does not explicitly state the classifying, with the machine learning model, the defect, including determining a type of the defect.
Narasimhan et al is analogous art pertinent to the technological problem addressed in this application and teaches the classifying, with the machine learning model, the defect, including determining a type of the defect (defects detected on the specimen in the image are determined to generate a defect map of the wafer maps and classified based on the type of defect using a deep learning model; Fig 1-3, 7 and ¶ [0039]-[0042], [0068]-[0069]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Cohen et al with Narasimhan et al including the classifying, with the machine learning model, the defect, including determining a type of the defect.
Regarding Claim 2, Cohen et al in view of Narasimhan et al teach the method of claim 1 (as described above), wherein modifying the subset of the potential defect indicators is additionally based on a number of the potential defect indicators in the arrangement (Cohen et al, a predetermined number of defect candidates having a selection threshold or parameter may be selected 204 as the subset to modify 206; Fig 2 and ¶ [0067]-[0068]).
Regarding Claim 3, Cohen et al in view of Narasimhan et al teach the method of claim 1 (as described above), wherein modifying the subset of the potential defect indicators is additionally based on a spatial distribution of the potential defect indicators in the arrangement (Cohen et al, the subset of defect candidates may also be based on a spatial attribute representing a specific pattern; Fig 2 and ¶ [0067]).
Regarding Claim 4, Cohen et al in view of Narasimhan et al teach the method of claim 1 (as described above), wherein modifying the subset of the potential defect indicators is additionally based on a skewness of the arrangement of the potential defect indicators (Cohen et al, a polynomial relation between pixels of an inspection image patch and corresponding reference image patch can be generated that identifies a polynomial shift 404 from variation (noise estimation 212), which is reflective in a gradient pattern with defects positions with a spatial anomaly (skewness) 507 on the pattern polynomial 503; Fig 2, 4, 5 and ¶ [0079]-[0083]).
Regarding Claim 5, Cohen et al in view of Narasimhan et al teach the method of claim 1 (as described above), further comprising identifying a cluster of some of the potential defect indicators that caused the defect (Cohen et al, the attributes of the defects detected in the defect map based on attribute space may be grouped in clusters based on machine learning techniques; ¶ [0102]-[0103]).
Regarding Claim 8, Cohen et al in view of Narasimhan et al teach the method of claim 1 (as described above), wherein generating the defect map includes combining a plurality of article defect maps (Narasimhan et al, the wafer map can be grouped with similar wafer maps; ¶ [0048]), each of the plurality of article defect maps corresponding to a different manufactured article (wafer maps are groups based on similar classification (of defect) and represent different signatures (different manufactured articles); Fig 1, 3 and ¶ [0045]-[0048]).
Regarding Claim 9, Cohen et al teach a computer system for classifying a defect in a manufactured article (system 100 to execute several modules for examining a specimen via an image to detect a defect; Fig 1, 2 and ¶ [0051]), comprising:
one or more processors (processing unit 102; Fig 1 and ¶ [0051]); and
non-transitory computer readable storage media encoding instructions (non-transitory computer readable memory, comprised in the processing unit 102, that stores instructions; Fig 1 and ¶ [0051]) which, when executed by the one or more processors (instructions are executed on the processing unit 102; Fig 1 and ¶ [0051]), causes the computer system to: perform steps identical to claim 1 (as described above).
Regarding Claim 10, Cohen et al in view of Narasimhan et al teach the computer system of claim 9 (as described above), with further claim limitations identical to claim 2 (as described above).
Regarding Claim 11, Cohen et al in view of Narasimhan et al teach the computer system of claim 9 (as described above), with further claim limitations identical to claim 3 (as described above).
Regarding Claim 12, Cohen et al in view of Narasimhan et al teach the computer system of claim 9 (as described above), with further claim limitations identical to claim 4 (as described above).
Regarding Claim 13, Cohen et al in view of Narasimhan et al teach the computer system of claim 9 (as described above), with further claim limitations identical to claim 5 (as described above).
Regarding Claim 14, Cohen et al in view of Narasimhan et al teach the system of claim 9 (as described above), wherein the machine learning model (Narasimhan et al, defects detected and classified based on the defect type using a deep learning model; Fig 1-3, 7 and ¶ [0039]-[0045], [0058]- [0064], [0068]-[0069]): receives the defect map in a neural network (Narasimhan et al, the deep learning model receives a feature map representing the specimen image with defects detected; ¶ [0068]-[0069]), wherein the neural network comprises a plurality of layers of neurons, including an input layer, one or more hidden layers, and an output layer (Narasimhan et al, the deep learning model includes one or more fully connected layers (FCL) with multiple nodes (neurons) and the FCL include input layers, processing (hidden) layers and output layers; ¶ [0058], [0062]-[0063] [0068]); ensures each neuron in each layer is connected to each neuron of a subsequent layer, and the output layer is connected to each neuron of a previous hidden layer of the one or more hidden layers (Narasimhan et al, in FCL, the nodes are connected between each of the layers; ¶ [0058], [0062]-[0063], [0068]); assigns a bias value to each neuron in the neural network (Narasimhan et al, a bias value can be assigned to the node; ¶ [0051]); and assigns a weight value to each connection in the neural network (Narasimhan et al, a weight can be assigned to the given neuron; ¶ [0062]).
Regarding Claim 15, Cohen et al in view of Narasimhan et al teach the computer system of claim 14 (as described above), wherein the output layer is configured to produce an image in which at least a portion of one or more of the potential defect indicators are collected into one or more defect groupings (Narasimhan et al, the output from the deep learning model can be an image that includes the defect and the classification and are grouped based on classification; Fig 2 and ¶ [0068]-[0069]).
Regarding Claim 16, Cohen et al in view of Narasimhan et al teach the computer system of claim 15 (as described above), wherein one or more of the potential defect indicators collected into one or more defect groupings are fit to at least one of a line, a curve, or a shape defined by a closed loop path (specification ¶ [0062] describes “closed loop path” and it is noted the groupings are fit to “at least one of…or”) (Narasimhan et al, the image defect detection classification may include a defect description, such as a pattern or bridge; Fig 2 and ¶ [0069]).
Claims 6, 17 are rejected under 35 U.S.C. 103 as being unpatentable over Cohen et al (US 2020/0234417) in view of Narasimhan et al (US 2022/0067911) and Yuk et al (Feature-Learning-Based Printed Circuit Board Inspection via Speeded-Up Robust Features and Random Forest).
Regarding Claim 6, Cohen et al in view of Narasimhan et al teach the method of claim 1 (as described above).
Cohen et al in view of Narasimhan et al do not teach calculating a kernel density of each potential defect indicator.
Yuk et al is analogous art pertinent to the technological problem addressed in this application and teaches calculating a kernel density of each potential defect indicator (a weighted kernel density estimation (WKDE) map is generated along with a WKDE map representing the probabilities of the pattern features of the manufactured good (a printed circuit board); Fig 4, 7, 8 and 3.3 Generation of WKDE Map by Weighting the Probability ¶ 1-6).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Cohen et al in view of Narasimhan et al with Rundo et al including calculating a kernel density of each potential defect indicator. By calculating a kernel density of the features, including defective areas, the accuracy of inspecting a manufactured good for defects is improved, thereby leading to increased quality in manufacturing, as recognized by Yuk et al (1. Introduction ¶ 5-6, 9).
Regarding Claim 17, Cohen et al in view of Narasimhan et al teach the computer system of claim 16 (as described above).
Cohen et al in view of Narasimhan et al do not teach wherein the line or the curve is plotted on the image to indicate a general shape or form of the one or more defect indicators collected into the one or more defect groupings.
Yuk et al is analogous art pertinent to the technological problem addressed in this application and teaches wherein the line or the curve is plotted on the image to indicate a general shape or form of the one or more defect indicators collected into the one or more defect groupings (the resultant image after extracting features include defining a lozenge area (line shape to distinguish area) so the fault (defect) area can be visually represented in the clearest manner; Fig 6 and 3.1 Speeded-up robust features (SURF) Based Feature Extraction and Class Assignment; ¶ 2).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Cohen et al in view of Narasimhan et al with Rundo et al including wherein the line or the curve is plotted on the image to indicate a general shape or form of the one or more defect indicators collected into the one or more defect groupings. By defining the given area of interest with a line shape outline, the fault area can be clearly visually represented, thereby proving a easily seen representation of the defect, as recognized by Yuk et al (3.1 Speeded-up robust features (SURF) Based Feature Extraction and Class Assignment; ¶ 2).
Claims 7, 18 are rejected under 35 U.S.C. 103 as being unpatentable over Cohen et al (US 2020/0234417) in view of Narasimhan et al (US 2022/0067911) and Rundo et al (US 2024/0202908).
Regarding Claim 7, Cohen et al in view of Narasimhan et al teach the method of claim 1 (as described above),
Cohen et al in view of Narasimhan et al do not teach wherein the at least one spatial attribute is an adaptive distance unique to each defect map.
Rundo et al is analogous art pertinent to the technological problem addressed in this application and teaches at least one spatial attribute is an adaptive distance unique to each defect map (the wafer defect patterns are determined using a distance-based similarity measure as compared to a relative distance per the defect pattern and a given threshold (thereby adaptive); Fig 3 and ¶ [0069]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Cohen et al in view of Narasimhan et al with Rundo et al including at least one spatial attribute is an adaptive distance unique to each defect map. By using a distance analysis of related datapoints and a threshold to determine similarity, defects may be identified and classified based on a particular defect pattern, thereby improving the inspection process during manufacturing, resulting in improved identification of defects to address for improved fabrication quality, as recognized by Rundo et al (¶ [0057]-[0059]).
Regarding Claim 18, Cohen et al in view of Narasimhan et al teach the computer system of claim 14 (as described above).
Cohen et al in view of Narasimhan et al do not teach wherein the neural network includes a plurality of channels, wherein each channel includes a plurality of layers of neurons, including an input layer and one or more hidden layers, wherein each neuron in each layer is connected to each neuron of a subsequent layer within each channel, and wherein each neuron in each layer is connected to each neuron of the subsequent layer within each of the plurality of channels.
Rundo et al is analogous art pertinent to the technological problem addressed in this application and teaches wherein the neural network includes a plurality of channels (the CNN comprises input layers as a plurality of data channels (example of at 224x224x3); Fig 12 and ¶ [0125]), wherein each channel includes a plurality of layers of neurons (each layer is characterized by a plurality of neurons; Fig 12 and ¶ [0125]), including an input layer and one or more hidden layers (the CNN includes an input layer 1204 to a plurality of hidden layers 1206-1222; Fig 12 and ¶ [0125]-[0126]), wherein each neuron in each layer is connected to each neuron of a subsequent layer within each channel (each layer of the fully connected layers connect each of the neurons between the layers; Fig 12 and ¶ [0125]-[0126]), and wherein each neuron in each layer is connected to each neuron of the subsequent layer within each of the plurality of channels (the neurons are fully connected layers 1220, 1222 for the given classes; Fig 12 and ¶ [0125]-[0126]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Cohen et al in view of Narasimhan et al with Rundo et al including wherein the neural network includes a plurality of channels, wherein each channel includes a plurality of layers of neurons, including an input layer and one or more hidden layers, wherein each neuron in each layer is connected to each neuron of a subsequent layer within each channel, and wherein each neuron in each layer is connected to each neuron of the subsequent layer within each of the plurality of channels. By using fully connected layers, features may be analyzed to determine if the feature correlates to multiple classes, thereby improving the analysis in classification of the identified features, as recognized by Rundo et al (¶ [0126]).
Claims 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Rundo et al (US 2024/0202908) in view of Yang et al (A robust vision inspection system for detecting surface defects of film capacitors).
Regarding Claim 19, Rundo et al teach a method of grouping two or more defect indicators on a surface of a manufactured article into a defect grouping (method of using system 300 to analyze wafer defect map images to identify, predict and classify wafer defect patterns; Fig 3 and ¶ [0068]-[0069]), the method comprising:
comparing a distance between a first defect indicator and a second defect indicator to a (the wafer defect patterns are determined using a distance-based similarity measure as compared to a relative distance per the defect pattern and a given threshold (thereby adaptive); Fig 3 and ¶ [0069]).
Rundo et al does not teach a dynamic threshold, wherein the dynamic threshold is a function of a distribution of at least the first defect indicator and the second defect indicator across the surface of the manufactured article.
Yang et al is analogous art pertinent to the technological problem addressed in this application and teaches a dynamic threshold, wherein the dynamic threshold is a function of a distribution of at least the first defect indicator and the second defect indicator across the surface of the manufactured article (defects based on surface characterization of capacitors during manufacturing process and compared to an adaptive (dynamic) threshold; Fig 2, 3, 7, 8 and 1. Introduction ¶ 3, 4.2 NSCT based inspection ¶ 8).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Rundo et al with Yang et al including a dynamic threshold, wherein the dynamic threshold is a function of a distribution of at least the first defect indicator and the second defect indicator across the surface of the manufactured article. By using an adaptive threshold, noise and false edges are removed from the defect detection analysis, thereby improving the detection of true defects during manufacturing and improves the speed, accuracy and quality in producing the manufactured article, as recognized by Yang et al (1. Introduction ¶ 3).
Regarding Claim 20, Rundo et al in view of Yang et al teach the method of claim 19 (as described above), wherein as the distance between the first defect indicator and the second defect indicator increases, a probability of merging the first defect indicator and the second defect indicator into a single defect grouping decreases (Rundo et al, groupings of wafer defect patterns identified as less similar, such as outside a similarity measure threshold, are determined to not be similar and would not be grouped in the same defect pattern cluster or class; Fig 3 and ¶ [0069]-[0070]).
Allowable Subject Matter
Claim 21 is objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Claim 21 is additionally objected to for clarity of the equation as claimed (discussed above in Claim Objections) but is interpreted as equation (1) of applicant’s specification (specification ¶ [0064]). However, the limitations of claim 21, in combination with the claims in which it depends were not readily found in the searched prior art, including the following limitation(s):
The method of claim 19, wherein the dynamic threshold is computed according to a following function:
PNG
media_image1.png
26
58
media_image1.png
Greyscale
wherein d, denotes the distance between the first defect indicator and the second defect indicator, I is a normal distribution distance of the first defect indicator and the second defect indicator, Sb is a kernel density of the first defect indicator, S, is a kernel density of the second defect indicator, and a is a factor to control a density effect.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Zhong et al (Weighted Matrix Decomposition for Small Surface Defect Detection) teach a weighted matrix decomposition model for defect detection on complex surfaces during machining processes, including use of surface mapping to detect defects.
Milligan et al (US 2018/0330493) teach a method and system to apply machine learning for defect pattern detection during semiconductor manufacturing including generation of defect variant patterns superimposed on wafer maps.
Kulkarni et al (US 5991699) teach techniques for improving manufacturing processes of semiconductors including defect detection by comparing the given feature representation to an expected threshold and grouping of detected defects.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KATHLEEN M BROUGHTON whose telephone number is (571)270-7380. The examiner can normally be reached Monday-Friday 8:00-5:00.
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, John Villecco can be reached at (571) 272-7319. 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.
/KATHLEEN M BROUGHTON/Primary Examiner, Art Unit 2661