A fully automatic point cloud data registration algorithm was proposed to disorderly point cloud with no additional information other than coordinates of measured points.In the algorithm,the normal vector and curvature were first estimated according to its neighbor points and Least-Squares approximation. Secondly,the EGI and CEGI were calculated.Then, the calculation of Euler angles of rotation through spherical harmonic functions and the translation according to Fourier transform were completed.With the initialization of the former results,the iterative closest point algorithm leads to perfect registration.Experimental results indicated that this kind of registration algorithm has a better effect.
