In order to solve the problem of registration of spatial objects with unobvious same-name feature points in the scanning at different sites or adjacent 3D field scenic spots, a point cloud registration method based on the center of gravity feature transformation is proposed. Firstly, the gravity centers of the point sets are calculated for the target point cloud and the point cloud to be registered, respectively. And then, the distance from the point set to the gravity center is calculated and arranged in ascending sort. According to the closest point and the farthest point, two feature vectors are constructed and the third feature vector is synthesized. Thus, the point set features of the two scenes can be described as three corresponding feature vectors. The rotation matrix is solved based on the rotation invariance of the matrix so as to solve the translation matrix and accomplish the initial registration. Finally, precise registration of the point cloud is done using the improved ICP algorithm. The proposed algorithm is analyzed experimentally with a plurality of registration algorithms, and the results show that the proposed algorithm is superior to the classic ICP algorithm in registration speed and precision, and highly practicable since it can effectively solve the registration problem of space targets with unobvious same-name feature points in the matching of point clouds at adjacent 3D scenes.