Method and device for on-line detection of paper product defects based on machine vision
By combining multi-line array cameras and YOLO networks, frequency domain residual perturbation analysis and multi-scale feature fusion of paper product defects are performed, solving the problem of low defect identification accuracy in high-speed production and realizing refined quality grading and online quality reporting of paper products.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- WUXI LANYAN PACKAGE CO LTD
- Filing Date
- 2026-03-30
- Publication Date
- 2026-07-17
AI Technical Summary
Existing online inspection technologies for paper products suffer from low accuracy in defect identification due to complex image interference and weak defect features during high-speed continuous production, making it difficult to achieve refined quality grading management.
Image acquisition is performed using a multi-line array camera. Multi-scale feature fusion is combined with frequency domain residual perturbation analysis and YOLO network to generate suspected defect mask images. Real-time hierarchical labeling is performed through a dynamic defect dictionary to generate online quality reports.
It improves the accuracy of defect identification in paper products, enables refined classification and quantitative analysis of defects, and supports online quality grading and process correlation mining.
Smart Images
Figure CN122415455A_ABST