PCB defect detection based on YOLOv10 algorithm with multi-scale mixed attention improvement mechanism

With the development of electronic information technology, PCB quality is crucial to the stability and reliability of electronic equipment, and its defect detection has become a key challenge in the electronics manufacturing industry. There are many drawbacks in traditional detection methods, and the rise of deep learning technology brings new opportunities for PCB defect detection. This study is based on the improved YOLOv10n model, in which the PSA layer is replaced by the multi-scale Mixed attention layer (MSSDSA), and trained on a dataset containing multiple PCB defects. The experimental results show that the loss value of the improved model decreases, and the mAP95 index increases from 0.630 to 0.636, which can detect the small and complex defects on PCB more accurately, and effectively overcome the problems such as large difference in defect size and strong background interference on PCB board while maintaining the real-time detection speed. It provides a high-precision automation solution for industrial quality inspection field, expands the application of attention mechanism in target detection, and has important value for the development of intelligent quality inspection system in electronic manufacturing industry.