GAN-Based Surface Defect Detection Reducing Overkill Rate
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting surface defects in industrial articles often result in a high overkill rate, where normal articles are mistakenly recognized as defective due to complex textures, leading to inefficient and inaccurate defect detection.
Innovation Solution
A method utilizing a generative adversarial networks model for defect detection, which includes a defect detection model and an image discrimination model, where the image discrimination model is trained using defect-free articles to adjust the defect detection results and reduce false positives by determining whether the surface image is truly defective based on reconstruction errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automatic defect detection is employed, then detection efficiency is improved, but overkill rate increases due to complex texture recognition
Solution Approach 1:
A generative adversarial network model is introduced as an intermediary between the defect detection model and the final detection result. The GAN model includes a generator that reconstructs surface images and a discriminator that verifies whether reconstructed images contain defects, serving as a mediator to filter false positives from the initial detection
Solution Approach 2:
The system implements a feedback mechanism where the defect detection model's results are fed into the GAN model for verification, and the GAN's determination results are used to adjust the original detection results. This closed-loop feedback process continuously refines the detection accuracy by using the discriminator's verification feedback to correct false positives
2Reliability
If manual recognition is used, then overkill rate is reduced, but detection efficiency decreases
Solution Approach 1:
The patent replaces manual mechanical recognition with an automated multi-stage detection system. The first stage uses an automated defect detection model for high-speed initial screening, and the second stage employs a GAN-based verification system to automatically filter false positives, substituting manual inspection entirely while maintaining low overkill rates
Solution Approach 2:
The detection process is segmented into two distinct stages: (1) initial defect detection using a defect detection model that provides high-speed automated screening, and (2) verification stage using the GAN model's discriminator to specifically verify suspected defect areas. This segmentation allows each stage to specialize in its function, maintaining both efficiency and accuracy
Data Source
AI summary
A method for detecting a surface defect, a method for training model, an apparatus, a device, and a medium, are provided. The method includes: inputting a surface image of the article for detection into a defect detection model to perform a defect detection, and acquiring a defect detection result output by the defect detection model; inputting a surface image of a defective article determined to be defective into an image discrimination model based on the defect detection result to determine whether the surface image of the defective article is defective, wherein the image discrimination model is a trained generative adversarial networks model, and the generative adversarial networks model is obtained by training using a surface image of a defect-free good article; and adjusting the defect detection result of the surface image of the defective article according to a determination result of the image discrimination model.