Automated Image Classification for Low-Contrast Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing inspection methods face challenges in accurately classifying articles due to issues with evaluation method determination, particularly when dealing with images of low-contrast defects or small defects, and require user intervention, making the process complex and time-consuming.
Innovation Solution
A method that automatically determines an evaluation method by creating a feature list from learning images, adjusts the evaluation method to increase dissimilarity between groups by modifying the evaluation values of singular images, and classifies articles based on the updated evaluation method, reducing user intervention and improving classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the evaluation method is determined by learning using multiple learning images, then the classification performance is improved, but the evaluation method may not be determined accurately when learning images include images with small defects or low-contrast defects
Solution Approach 1:
The system performs preliminary classification of learning images into multiple groups based on defect characteristics before determining the evaluation method. This preliminary action allows the system to identify images with small defects or low-contrast defects and handle them appropriately during the learning process, ensuring the evaluation method is determined accurately for all defect types.
Solution Approach 2:
The system changes parameters of the evaluation method based on the group to which the learning image belongs. By adjusting evaluation parameters according to defect characteristics (such as defect size, contrast level, or type), the system can accurately evaluate diverse defect conditions that would otherwise be misclassified.
2Measurement precision
If the user is prompted to change the evaluation method based on feature amount distribution, then the classification accuracy can be improved, but the process becomes complex and requires considerable user skill and time
Solution Approach 1:
The system automatically determines and adjusts the evaluation method without requiring user intervention. The classification unit autonomously performs feature amount distribution analysis, identifies appropriate evaluation methods, and updates the learning model, thereby simplifying the learning process and eliminating the need for user skill in method selection.
Solution Approach 2:
The system uses feedback from the classification results and feature amount distributions to automatically adjust the evaluation method. By continuously monitoring classification performance and adjusting parameters based on this feedback, the system improves accuracy while maintaining operational simplicity.
3Measurement precision
If all samples are classified every time the evaluation method is changed, then the classification performance can be optimized, but the learning process becomes complicated and time-consuming
Solution Approach 1:
The system performs preliminary classification of learning images into multiple groups based on defect characteristics before determining the evaluation method. This preliminary organization allows the system to efficiently update the evaluation method by focusing on relevant sample groups rather than reclassifying all samples, thereby reducing learning time while maintaining optimization effectiveness.
4Measurement precision
If the number of selectable image features increases to improve classification accuracy, then the evaluation method becomes more comprehensive, but determination by the user becomes more difficult and time-consuming
Solution Approach 1:
The system automatically selects and determines the appropriate image features and evaluation method without user intervention. The classification unit autonomously analyzes feature amount distributions, identifies the most relevant features for each defect group, and configures the evaluation method accordingly, making the process easy to operate despite the large number of available features.
Data Source
AI summary
The present invention provides a classifying method of classifying an article into one of a plurality of groups based on an image of the article, comprising determining an evaluation method for obtaining an evaluation value of an image by using at least some of sample images, obtaining evaluation values for the sample images by the determined evaluation method, changing the evaluation method so as to increase a degree of dissimilarity in an evaluation value range for sample images between the plurality of groups by changing a evaluation value of at least one sample image having a singular evaluation value among the sample images, obtaining an evaluation value for the image of the article using the changed evaluation method, and classifying the article into one of the plurality of groups based on the evaluation value for the image of the article.