Learning Model Virtual Good Article Image Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing appearance inspection methods require advanced knowledge and experience to determine feature values and determination standards for defect detection, making them difficult to develop and implement effectively.
Innovation Solution
An inspection apparatus using a learning model to convert inspection target images into virtual good article images, generating defect candidate images by comparing these with the original images, and displaying them for user input to restructure the learning model and exclude over-detected defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image processing methods are used for defect detection, then defect detection capability is achieved, but advanced knowledge and experience are required to determine feature values and determination standards
Solution Approach 1:
The learning model automatically learns and determines appropriate feature values and determination standards from training data without requiring manual configuration by experts. The system performs self-adjustment through machine learning, eliminating the need for advanced knowledge in developing the inspection system while maintaining high defect detection capability
Solution Approach 2:
The system transforms the inspection approach by changing from manual parameter setting to automated parameter learning. The learning model dynamically determines optimal feature values and determination standards based on training data, converting a complex manual configuration process into an automated parameter optimization process
2Reliability
If manual determination of feature values and standards is performed, then accurate defect detection is achieved, but the development process becomes difficult and time-consuming
Solution Approach 1:
The system performs preliminary learning by training the model with training data before actual inspection begins. This preliminary action of automated learning replaces the time-consuming manual determination process, allowing the system to quickly adapt to specific inspection requirements without requiring extensive manual configuration time
Solution Approach 2:
The learning model uses feedback from training data to automatically adjust and optimize feature values and determination standards. This automated feedback mechanism eliminates the need for manual iterative adjustment, significantly reducing development time while maintaining or improving detection accuracy through continuous optimization
Data Source
AI summary
An inspection apparatus including: a display device; and one or a plurality of processors, whereinthe one or the plurality of processors is programmed to execute a method including: converting an inspection target image representing an inspection target into a virtual good article image by using a learning model, the learning model being trained so that an image representing a good article is generated based on features of a plurality of targets that are determined as good articles, generating a difference between the virtual good article image and the inspection target image as a defect candidate image, and displaying the defect candidate image on the display device.