Film defect detection method, device, apparatus, product, and storage medium

By generating diverse defect samples using a generator and training the model using a meta-learning strategy, the problems of sample scarcity and texture variation in membrane material defect detection are solved, achieving high-precision and fast-adaptive detection results while reducing data preparation costs.

CN122434856APending Publication Date: 2026-07-21SHENZHEN INSTITUTE OF INFORMATION TECHNOLOGY
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SHENZHEN INSTITUTE OF INFORMATION TECHNOLOGY
Filing Date
2026-04-22
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

In membrane material defect detection, existing technologies rely on a large amount of high-quality labeled data, especially since rare defect samples are scarce, resulting in low detection accuracy and recall. Furthermore, the models struggle to adapt quickly to texture changes and novel defects, increasing deployment and maintenance costs.

Method used

A diverse range of defect samples are generated by a generator, and a pre-defined membrane defect detection model is trained using a meta-learning strategy. An attention mechanism is used to decouple defect and background features, and data augmentation and loss function optimization are performed to construct a high-precision membrane defect detection model.

Benefits of technology

Achieving high-precision membrane defect detection under limited sample conditions reduces reliance on real samples, improves the model's generalization ability and adaptability, and lowers data acquisition and annotation costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122434856A_ABST
    Figure CN122434856A_ABST
Patent Text Reader

Abstract

The application discloses a film defect detection method, device, equipment, product and storage medium, and relates to the technical field of industrial visual detection. The method comprises the following steps: in response to a film defect detection instruction, a film image is acquired, a preset film defect detection model is used to process the film image, and a film defect detection result is obtained. The preset film defect detection model is obtained based on diversified defect samples after training. The diversified defect samples are generated by a generator based on defect morphological characteristics corresponding to real film material images. In a few-sample scene, a large number of diversified defect samples are generated by the generator based on defect characteristics corresponding to a small number of real film material images. The model is trained based on the large number of diversified defect samples, the accuracy of the output result of the film defect detection model is ensured, and high-precision film defect detection is realized.
Need to check novelty before this filing date? Find Prior Art