Study on the PCB Defect Detection Method Based on CPH-YOLOv5

The current PCB (Printed Circuit Board) defect detection algorithm can meet certain practical needs, but there are still problems such as small defects that are difficult to identify and diverse defect shapes that lead to a decrease in recognition rate. Therefore, a CPH-YOLOv5 (ConvNeXt Prediction head, CPH) PCB defect detection algorithm for detecting small defects in PCBs is proposed. Firstly, adding a small target detection layer and removing the original large target detection layer, so as to improve the model's ability to perceive small target features. Secondly, decoupling the original coupling header, extracting the target location and category information separately, learning through different network branches, finally fusing, thus reducing the number of parameters and computational complexity and enhancing the generalization ability and robustness of the mode. At last, introducing the ConvNeXt structure. It can enable the model to obtain richer background and global information, which leads to a higher accuracy of the model. The experimental results show that the accuracy of the improved model can reach 99.32%, which is 0.66% higher than that of the original YOLOv5 algorithm, and can meet the requirements of PCB defect detection.