A Tool for Printed Circuit Board Defect Detection based on the YOLO Algorithm

Printed circuit board (PCB) is an essential component of modern machines. To solve the problems of a wide variety of PCB defects with complex shapes, small defect regions, and difficulties in accurately locating defects, a tool for PCB defect detection based on the You Only Look Once (YOLO) algorithm is implemented. Six types of PCB defects, i.e., Missing-hole, Mouse-bite, Open-circuit, Short, Spur, and Spurious-copper, can be detected through the designed tool. The attention mechanism and multi-scale feature fusion structure are introduced to the original YOLOv5s algorithm to decrease the missed detection rate of small defects and improve the detection speed. An intelligent defect detection system is then designed and implemented to meet the real-time detection requirements in actual PCB production scenario. The experimental results show that the detection accuracy and F1 score of the improved YOLOv5s system reach 97.4% and 0.97, respectively.