Defect Classifier Using Distribution Data for Manufacturing Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated defect categorization in manufacturing industries faces challenges in accurately determining the category of defects, especially when confidence levels are approximate, leading to incorrect categorization due to similar probabilities across categories.
Innovation Solution
A method that utilizes defect distribution information from training images to determine the target category of a defect by applying the image and defect distribution data to a defect classifier, considering the frequencies of defects at corresponding locations, thereby improving categorization accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated visual inspection technologies are used for defect categorization, then productivity is improved, but measurement precision deteriorates when confidence levels are approximate
Solution Approach 1:
The patent introduces defect distribution information as an intermediary element that mediates between the defect classifier's confidence levels and the final categorization decision. This intermediary provides additional contextual information about defect frequencies at specific locations, enabling more accurate categorization when confidence levels are approximate or ambiguous
Solution Approach 2:
The patent changes the parameter set used for categorization by incorporating defect distribution information (frequencies of defects at corresponding locations) alongside the classifier's confidence levels. This parameter expansion allows the system to resolve ambiguities that would otherwise lead to incorrect categorization
2Device complexity
If defect categorization relies solely on classifier confidence levels, then device complexity is reduced, but reliability deteriorates due to incorrect categorization
Solution Approach 1:
The patent performs preliminary action by pre-computing and storing defect distribution information from training images before the actual defect categorization task. This pre-computed information is then available to guide categorization decisions, improving reliability without adding complexity to the real-time classification process
Solution Approach 2:
The system uses defect distribution information as a feedback mechanism that provides additional constraints and guidance to the categorization process. When confidence levels are approximate, the feedback from defect distribution patterns helps correct potential misclassifications, thereby improving reliability
Data Source
AI summary
According to embodiments of the present invention, a method, a device and a computer program product for image processing is provided. A computing device obtains an image of a first object, the image presenting a defect of the first object. A computing device obtains defect distribution information indicating respective frequencies of a plurality of predetermined categories of defects presented at corresponding locations in a plurality of training images, the plurality of training images presenting second objects and being used for training a defect classifier. A computing device determines a target category of the defect of the first object by applying the image and the defect distribution information to the defect classifier. A computing device generates one or more correction notifications.