Research on PCB Defect Detection Method Based on Deep Learning

Due to the small size of defects on printed circuit board (PCB), traditional image processing methods often result in missed detections and high computational costs. This paper proposes a PCB defect detection algorithm based on an improved YOLOv8n model. In the backbone, the C2f module is combined with spatial and channel reconstruction convolution (SCConv) to reduce redundant features and significantly decrease model complexity. In the neck section, a context anchor attention (CAA) mechanism is introduced, which adaptively assigns higher weights to important features, enhancing the model's ability to capture small defects and effectively suppress background noise and irrelevant information. Experimental results on public datasets show that the proposed method achieves significant improvements in precision and mAP@0.5, with increases of 3.0% and 2.2%, respectively, compared to YOLOv8n. The proposed algorithm offers an efficient and practical solution for PCB defect detection, delivering high performance and significant application value.