A PCB bare board defects detection method based on dual-domain wavelet-YOLO

In this paper, a defect detection model DDW-YOLO based on YOLOv5 is proposed to improve the accuracy and efficiency of PCB bare board defect detection. In PCB bare boards, the ratio of defect area to background area differs significantly, resulting in poor performance of existing detection methods in identifying small target defects. To solve this problem, this paper proposes the wavelet signature fusion module and the dual-domain wavelet attention network, which can enhance the model's attention to high-frequency information and improve the detection ability of small defects. The experimental results show that DDW-YOLO is superior to the current mainstream model in both detection accuracy and calculation efficiency, and provides an effective solution for PCB bare board defect detection.