Deep learning method, system and medium for printed circuit board defect identification
By performing interlayer misalignment detection and feature extraction on X-ray scanning images of printed circuit boards, combined with via misalignment and pad offset analysis, a defect recognition model is constructed. This solves the problems of high data cost, large computing resource requirements and poor real-time performance in the existing technology for printed circuit board defect recognition, and achieves efficient and accurate defect recognition.
Patent Information
- Application Number
- CN202411735892.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-29
- Publication Date
- 2025-09-12
- Estimated Expiration
- 2044-11-29
AI Technical Summary
Existing deep learning models for printed circuit board defect recognition have problems such as high training data cost, large computing resource requirements, poor real-time performance, insufficient robustness, and limited ability to recognize circuit boards of different types and sizes. In particular, the recognition results are inaccurate when processing low-quality images or those containing noise.
By acquiring X-ray scanning images of printed circuit boards, interlayer misalignment detection and feature extraction are performed. Combined with via misalignment and pad offset analysis, binary conversion and solder joint area identification are performed, a defect recognition model is constructed, and modeling is performed using multi-source data to achieve automated defect recognition for circuit boards of different types and sizes.
It improves the accuracy and robustness of defect identification, reduces manual participation, improves detection efficiency and real-time performance, is applicable to circuit boards of different types and sizes, reduces detection costs, and enhances the system's real-time performance and breadth of application.