A plating layer appearance defect detection method based on big data processing
By combining progressive deformable convolutional networks with pseudo-label expansion mechanisms, the problem of inaccurate defect identification in electroplating appearance inspection is solved, achieving efficient and accurate detection of electroplating defects, and is suitable for quality assessment of electroplating under complex conditions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DONGGUAN LIHENG COATING TECH CO LTD
- Filing Date
- 2026-06-09
- Publication Date
- 2026-07-17
AI Technical Summary
Existing methods for inspecting the appearance of electroplated layers suffer from problems such as large subjective differences, high rates of missed detection of small defects, and low utilization of unlabeled samples. In particular, under conditions of complex reflective surfaces and local texture interference, it is difficult to accurately identify defects in electroplated layers.
A progressively deformable convolutional network structure and a pseudo-label expansion mechanism are adopted to perform hierarchical offset convolution calculation, confidence screening and iterative update on electroplating layer image samples to construct a continuously optimized defect recognition process. Pseudo-labels are generated using unlabeled samples to expand the training set, thereby realizing defect feature extraction and location determination.
It achieves sufficient extraction of electroplating layer defect features and accurate determination of defect location, making it suitable for electroplating inspection scenarios with limited sample annotation quantity. It outputs defect category, location, and area results, facilitating electroplating layer quality judgment and re-inspection positioning.
Smart Images
Figure CN122415592A_ABST