In view of the problem that the commonly used registration algorithms cannot meet the high-precision process requirements in the manufacturing industry, this paper proposes an improved 3DSC point cloud registration based on the 3D point cloud. Firstly, the threshold value is set to collect the contour point cloud by the improved down-sampling method, and the collected point cloud is divided into 3D mesh in turn to form a 3D shape context. Then, the improved 3DSC coarse registration is performed, and the ICP fine registration is used to realize the rotation and translation transformation between the source point cloud and the target point cloud. In order to verify the effectiveness of the improved algorithm, the traditional 3DSC, FPFH-ICP, PFH-ICP and the improved algorithm in this paper are used to compare the registration experiments. The experiment results show that the accuracy of the improved algorithm can reach 2.253 55e-05 m and 9.969 02e-06 m respectively for the bunny point cloud and the flowerpot point cloud, which is obviously better than the registration accuracy of other algorithms. Compared with the traditional 3DSC registration algorithm, the improved 3DSC registration algorithm can save 75%~85% of the registration time. The improved 3DSC point cloud registration method is beneficial to improve the registration accuracy, optimize the registration time, and improve the registration efficiency.