Abstract:Due to the particularity of the gear burr location and the similarity of the surrounding environment,the traditional image processing methods cannot achieve good results.Therefore,a target detection algorithm based on improved YOLO v3 network was proposed to realize rapid detection of gear burrs characteristics.By improving the resolution of the network input and adjusting the network structure,the performance of the improved YOLO v3 network was further optimized and the detection efficiency was improved.Before making labels,Zhangs calibration method was used to eliminate the influence of lens distortion on the pictures.The results show that compared with original YOLO v3 network,the improved YOLO v3 network has better detection results,its network size is reduced by 1/4,and the detection speed is increased by nearly 2 times.