Abstract: The key of point cloud registration is to build corresponding relation between two point clouds. Inheriting and optimizing algorithm is presented to realize precise point cloud registration. Algorithm principle is explained

and rotating parameters and shifting parameters can be obtained by building model. Realizing procedure of the algorithm is brought forward and analyzed. 6 groups of point clouds of car flashing are acquired by grating projection 3D scanning instrument

and point clouds registration based on presented algorithm is carried out. Registration error is analyzed by multi resolution hiberarchy method. It indicates that the proposed algorithm can realize precise registration of unorganized point clouds. Registration standard error is less than 0.10mm

and the registration time of 2 point clouds is less than 2 seconds. Compared to ICP algorithm

the required iterative number is reduced by more than 5 times.