Woven bag surface defect detection method based on deep learning
By using a multi-scale feature decoupling detection network and a texture suppression and defect enhancement coupling architecture, the problem of texture and defect overlap in the detection of woven bag surface defects is solved, thus improving the detection accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANDONG RIYUESHENG PACKAGING CO LTD
- Filing Date
- 2026-06-01
- Publication Date
- 2026-07-17
AI Technical Summary
Existing deep learning models struggle to distinguish between periodic textures and weak feature defects in the detection of surface defects on woven bags, leading to missed detections and false detections.
A multi-scale feature decoupling detection network is adopted. Periodic texture feature components and defect residual feature components are separated by periodic texture prior constraints. The feature weights are adjusted by a coupled architecture of texture suppression and defect enhancement to generate a target feature map with texture suppression and defect enhancement.
It effectively reduces the false negative and false positive rates of weak feature defects under the interference of periodic texture in woven bags, and improves the accuracy of defect detection.
Smart Images
Figure CN122416151A_ABST