基于改进光度立体的圆柱电池表面缺陷检测系统OA北大核心CSTPCD

To solve the problem of low efficiency and low precision of manual detection of cylindrical battery appearance,a cylindrical battery surface defect detection system based on improved photometric stereo is designed.Firstly,aiming at the problem of small gray difference of defects such as scratches and bumps in the surface image of the battery under single light source illumination,a four-direc-tion light source illumination device is designed to improve the contrast of the defect image.Secondly,according to the characteristics of significant changes in the depth of weak defects,a three-dimensional shape reconstruction method based on improved photometric stereo is proposed to quickly reconstruct the depth image of the battery surface.Combined with the direction difference of depth information,a multi-channel image fusion algorithm is proposed to extract and fuse depth features and surface curvature features to enhance the con-trast of weak defects.Finally,a multi-level fusion model based on YOLOv5 network is designed to complete the defect detection.The ex-perimental results show that the detection accuracy of the detection system for battery surface defects can reach 99.8%,and the stability and real-time performance are good.It has good feasibility in actual industrial scenarios.