# US 12,762,679 B2 · Defect characterization in semiconductor devices based on image processing > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,762,679 B2 - Document ID: US12762679B2 - Status: Granted - Kind: B2 - Application number: 18/388,623 - Filed: 2023-11-10 - Published: 2024-05-16 - Granted: 2026-10-13 - Earliest claimed: 2022-11-15 - Assignees: Micron Technology, Inc. - Applicants: Micron Technology, Inc. - Inventors: Nagasubramaniyan Chandrasekaran - Web page: https://patentgrep.com/patent/US12762679B2 ## CPC - G06T 7/0004: {Industrial image inspection} - G06V 10/25: Determination of region of interest [ROI] or a volume of interest [VOI] - G06V 10/40: Extraction of image or video features - G06V 10/764: using classification, e.g. of video objects - G06V 10/7715: {Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods} - G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting - G06T 2207/20081: Training; Learning - G06T 2207/30148: Semiconductor; IC; Wafer - G06V 20/50: Context or environment of the image - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract A system includes a memory and a processing device, operatively coupled with the memory, to perform operations including: receiving an image of a substrate of an electronic device; extracting, by a feature extraction model processing the image, a plurality of visual features from the image; and identifying, by a trainable feature classifier processing the plurality of visual features, a region of interest corresponding to an electronic circuit associated with performance of the electronic circuit. ## Claims ### Claim 1 A system comprising: a memory; and a processing device, operatively coupled with the memory, to perform operations comprising: receiving an image of a substrate of an electronic device, the image covering an area of the substrate, wherein a critical measurement of the substrate is calculated as an aggregated value over the area; determining a plurality of candidate regions in the image based on sampling information associated with the critical measurement, wherein the sampling information specifies a plurality of sampling locations within the area, and wherein values measured at the plurality of sampling locations are used to calculate the aggregated value; extracting, by a feature extraction model processing the image, a plurality of visual features from the image; and identifying, by a trainable feature classifier processing the plurality of visual features extracted by the feature extraction model, a region of interest among the plurality of candidate regions, the region of interest corresponding to an electronic circuit and associated with performance of the electronic circuit. ### Claim 2 (depends on 1) The system of claim 1, wherein the operations further comprise: in view of the region of interest, identifying a defect that leads to a failure of the electronic device. ### Claim 3 (depends on 1) The system of claim 1, wherein the operations further comprise: determining, by a trainable defect classification model processing a subset of the plurality of visual features associated with the region of interest, a type of a defect associated with the region of interest. ### Claim 4 (depends on 1) The system of claim 1, wherein the operations further comprise: receiving a second image of the region of interest, wherein a resolution of the second image exceeds a resolution of the image of the substrate; extracting, by a second feature extraction model processing the second image, a second plurality of visual features from the second image; and identifying, by a second trainable feature classifier processing the second plurality of visual features, a second region of interest corresponding to the electronic circuit associated with performance of the electronic circuit within the second image, wherein the second region of interest is a part of the region of interest. ### Claim 5 (depends on 4) The system of claim 4, wherein the operations further comprise: determining, by a trainable defect classification model processing a subset of the second plurality of visual features associated with the second region of interest, a type of a defect associated with the second region of interest. ### Claim 6 (depends on 1) The system of claim 1, wherein the feature extraction model is trainable. ### Claim 7 (depends on 1) The system of claim 1, wherein the feature extraction model implements at least one of: Scale Invariant Feature Transform (SIFT), Speeded Up Robust Features (SURF), KAZE, Accelerated-KAZE (AKAZE), Oriented FAST and Rotated BRIEF (ORB), or Binary Robust Invariant Scalable Keypoints (BRISK). ### Claim 8 (depends on 1) The system of claim 1, wherein the trainable feature classifier implements at least one of: Perceptron, Naive Bayes, Decision Tree, Logistic Regression, K-Nearest Neighbor, Artificial Neural Networks, Deep Learning, or Support Vector Machine. ### Claim 9 A non-transitory computer readable medium comprising instructions, which when executed by a processor, cause the processor to perform operations comprising: receiving an image of a semiconductor substrate of an electronic device, the image covering an area of the substrate, wherein a critical measurement of the substrate is calculated as an aggregated value over the area; determining a plurality of candidate regions in the image based on sampling information associated with the critical measurement, wherein the sampling information specifies a plurality of sampling locations within the area, and wherein values measured at the plurality of sampling locations are used to calculate the aggregated value; extracting, by a feature extraction model processing the image, a plurality of visual features from the image; and identifying, by a trainable feature classifier processing the plurality of visual features extracted by the feature extraction model, a region of interest among the plurality of candidate regions, the region of interest corresponding to an electronic circuit exhibiting suboptimal performance. ### Claim 10 (depends on 9) The non-transitory computer readable medium of claim 9, wherein the operations further comprise: preprocessing the image. ### Claim 11 (depends on 9) The non-transitory computer readable medium of claim 9, wherein the operations further comprise: determining, by a trainable defect classification model processing a subset of the plurality of visual features associated with the region of interest, a type of a defect associated with the region of interest. ### Claim 12 (depends on 9) The non-transitory computer readable medium of claim 9, wherein the operations further comprise: receiving a second image of the region of interest, wherein a resolution of the second image exceeds a resolution of the image of the semiconductor substrate; extracting, by a second feature extraction model processing the second image, a second plurality of visual features from the second image; and identifying, by a second trainable feature classifier processing the second plurality of visual features, a second region of interest corresponding to an electronic circuit exhibiting suboptimal performance within the second image, wherein the second region of interest is a part of the region of interest. ### Claim 13 (depends on 12) The non-transitory computer readable medium of claim 12, wherein the image is received from a first imaging device, and the second image is received from a second imaging device. ### Claim 14 (depends on 12) The non-transitory computer readable medium of claim 12, wherein the feature extraction model and the second feature extraction model use different feature detectors. ### Claim 15 (depends on 12) The non-transitory computer readable medium of claim 12, wherein the trainable feature classifier and the second trainable feature classifier are trained using different training data. ### Claim 16 (depends on 12) The non-transitory computer readable medium of claim 12, wherein the trainable feature classifier and the second trainable feature classifier are trained using different machine learning techniques. ### Claim 17 A method, comprising: receiving, by a processing device, an image of a substrate of an electronic device, the image covering an area of the substrate, wherein a critical measurement of the substrate is calculated as an aggregated value over the area; determining a plurality of candidate regions in the image based on sampling information associated with the critical measurement, wherein the sampling information specifies a plurality of sampling locations within the area, and wherein values measured at the plurality of sampling locations are used to calculate the aggregated value; extracting, by a feature extraction model processing the image, a plurality of visual features from the image; and identifying, by a trainable feature classifier processing the plurality of visual features extracted by the feature extraction model, a region of interest among the plurality of candidate regions, the region of interest corresponding to an electronic circuit associated with performance of the electronic circuit. ### Claim 18 (depends on 17) The method of claim 17, wherein the feature extraction model comprises at least one of: Scale Invariant Feature Transform (SIFT), Speeded Up Robust Features (SURF), KAZE, Accelerated-KAZE (AKAZE), Oriented FAST and Rotated BRIEF (ORB), or Binary Robust Invariant Scalable Keypoints BRISK). ### Claim 19 (depends on 17) The method of claim 17, wherein the trainable feature classifier comprises at least one of: Perceptron, Naive Bayes, Decision Tree, Logistic Regression, K-Nearest Neighbor, Artificial Neural Networks, Deep Learning, or Support Vector Machine.