As the automobile-bodies point cloud had the traits of large geometric dimension, huge data and rigor reverse precision, the improved iterative closet point algorithm (ICP) is put forward. The searching structure is generated using k-D tree. The closet points are searched by k-Dimensional sphere. The mapped relationship between points is generated. The matched points are filtered by the principle of geometric similarity. The solving of ICP algorithm is speeded and the registration precision of ICP algorithm is improved. The registration algorithm based on least-square and quaternion is used to calculate accurate registration result. The algorithm has certain theoretical and practical significance for improving of the efficiency and precision of registration. The reliability and accuracy of the algorithm are proved by experimentation.