Current PCB defect detection algorithms suffer from low detection accuracy and slow detection speed.To solve the probleman improved PCB defect detection algorithm based on YOLO v3 network is proposed.Firstlybased on DBSCAN+k-means clustering algorithmre-clustering is performed by using Avg IOU criteria to select the Anchor Boxes that are more suitable for the data set.Secondlytwo residual units are added to the second residual module to improve the networks ability to extract shallow features.At the same timeSE Block module is added to the network to highlight useful feature channels and improve the structure of feature fusion.Finallythe detection module is modified to improve the detection ability on the data set.Experimental results show that the improved algorithm significantly enhances detection accuracy and detection speed on PCB defect data set.