With the improvement of integrated circuit integration,the industrial complexity of printed circuit boards has gradually increased,and the defects in the manufacturing process have become more minor and complex.Aiming at the problem that the traditional image processing algorithm is difficult to accurately detect various minor defects,a defect detection and recognition algorithm for printed wiring board based on the improved YOLOv5s is proposed.Firstly,the three-scale detection layer of the original network is changed,and the original large target detection layer is replaced with a detection layer that detects smaller targets,so as to strengthen the detection accuracy of the model for small and medium-sized targets.At the same time,the K-means++algorithm is used to recluster the prior boxes.The PANet network is then adjusted to a BiFPN network,and the importance of each feature is evaluated using learnable weights to strengthen feature fusion.Finally,SENet is incorporated into the C3 module,allowing the network to selectively focus on information characteristics and filter redundant information.Final results indicated that the detection precision of the method of this article attain 97.83%and the recall rate is 95.73%.the*spur'false detection and‘mouse_bite'missed detection rate is greatly reduced compared with the original network,which completely satisfy the industrial detection requirements