一种玻璃盖板无监督小样本缺陷检测方法
By combining unsupervised learning, few-shot learning, and incremental learning techniques, and utilizing U-Net and Swin Transformer for feature extraction and fusion, an adaptive detection model was established. This solved the problem of defect detection in cover glass under unsupervised few-shot conditions, and achieved efficient and accurate automated detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHENGZHOU UNIV
- Filing Date
- 2024-01-09
- Publication Date
- 2026-07-17
AI Technical Summary
Existing deep learning models struggle to effectively detect small defects in cover glass under unsupervised, limited-sample conditions, and manual inspection is inefficient and unreliable, failing to meet the rapid and efficient demands of industrial production.
We employ unsupervised learning, few-shot learning, and incremental learning techniques, combining U-Net and Swin Transformer for feature extraction. Through feature alignment and fusion, we establish an adaptive, dynamically scalable detection model and perform incremental learning to optimize the detection model.
It improves the accuracy and efficiency of cover glass defect detection, solves the detection difficulties of unsupervised models under conditions of few samples, and realizes rapid and efficient automated detection.
Smart Images
Figure CN117830278B_ABST