Abstract:In the field of 3D modeling, the precision of the traditional ICP algorithm is not high when the point cloud overlap is low, so we propose a registration algorithm based on consistent sphere. For the registration algorithm, on finding the corresponding points, the proposed algorithm innovatively combines the rotation invariance of the sphere with the SVD orthogonal consistency algorithm based on neighborhood. This helps the registration algorithm obtain the corresponding points with higher accuracy, and on this basis, spread and get more corresponding points. It then uses the rigid constraints to eliminate the error points. Finally, we use the quaternion method to solve transformation matrix. The results show that the proposed algorithm has not only overcome the shortcomings of the traditional ICP algorithm, but has better accuracy.