Multimodal Defect Classification Using Image and Manufacturing Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing classification systems face challenges in accurately determining the class of objects, particularly in distinguishing between similar defects in products, as they rely solely on image data or manufacturing data, leading to uncertainty and reduced reliability in defect classification.
Innovation Solution
A classification system that combines image data and manufacturing data by using a learned model for image classification and a model-based approach for manufacturing data, allowing for the determination of object classes based on both classification results to enhance accuracy and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If only image data is used for classification, then the classification process is simple, but the reliability of class determination is reduced
Solution Approach 1:
The patent combines image data classification results with manufacturing data classification results in the determination part. This merging of multiple data sources and classification approaches resolves the contradiction by improving reliability through comprehensive analysis while managing complexity through systematic integration of the classification components.
2Measurement precision
If only image data is used for classification, then the processing load is low, but the accuracy of defect identification is reduced
Solution Approach 1:
The determination part merges image data classification results with manufacturing data classification results to achieve more accurate defect identification. This combination allows the system to leverage complementary information from both data sources, improving measurement precision while distributing the processing load across multiple specialized classification components.
3Reliability
If multiple classification results are combined, then the reliability of class determination is improved, but the device complexity increases
Solution Approach 1:
The patent segments the classification system into distinct functional parts: a first classification part for image data, a second classification part for manufacturing data, and a determination part for integrating results. This segmentation allows each component to specialize in specific tasks, improving reliability through comprehensive analysis while managing overall system complexity through modular architecture.
Solution Approach 2:
The determination part serves multiple functions: it receives classification results from both image and manufacturing data, decides whether to use one or both results, and determines the final class of the object. This multi-functionality resolves the contradiction by consolidating integration logic in a single component that handles multiple classification sources uniformly.
Data Source
AI summary
Reliability regarding a class determination for an object is improved. Classification system includes first classification part, second classification part, and determination part. First classification part classifies first target data into at least one of a plurality of first classes. Second classification part classifies second target data into at least one of a plurality of second classes. Determination part decides whether to use one or both of a first classification result that is a classification result obtained by first classification part and a second classification result that is a classification result obtained by second classification part, and determines a class of object based on one or both of them. The first target data is image data of object. The second target data is manufacturing data regarding a manufacturing condition of object.