Aiming at the problems of low accuracy and low efficiency in point cloud registration without any initial value, a rough registration method of 3D point cloud based on feature matching was proposed. Firstly, the local normal vector of the point cloud is used to describe its features, and feature retention weights are added to filter feature information to improve the efficiency of registration; Then feature histograms are built according to the reserved feature information, and the initial matching point pairs are obtained by comparing the information described by the feature histograms; Finally, the rigid invariant constraint and random sample consensus algorithm were combined to screen the correct matching point pairs, and the rotation matrix and translation vector were calculated by the quaternion method. Experimental results show that the proposed algorithm has higher accuracy and efficiency compared with other rough registration algorithms, and provides a good initial value for the subsequent fine point cloud registration.