An accurate and efficient algorithm of point cloud auto registration algorithm is proposed.Corresponding feature points in the images are found by using the SIFT algorithm

RANSAC algorithm is used to eliminate the outlier feature matches.The corresponding points in the point cloud is obtained

false matches are eliminated by using a voting scheme.The unit quaternion solution is employed for initial registration.An improved ICP algorithm based on feature points is used to achieve precise registration.Experiments show that this algorithm has fast and stable characteristics in point clouds registration.