Point cloud registration based on improved ICP algorithm

Point clouds have typically applications in environment perception and robot navigation, stereo vision, visual registration, and depth estimation. point cloud registration is one of the important steps in 3-D Point Cloud Processing when combining multiple point clouds to reconstruct a 3-D scene. The iterative closest point (ICP) algorithm is most commonly used when the point cloud data obtained from different perspectives are precisely registrated. The classical iterative closest point (ICP) algorithm converges slowly, In this paper, the voxel lattice is used to re-sample the point cloud data, and the kd tree is used to optimize the calculation of the normal vector. Instead, the method of finding the corresponding point is used to improve the accuracy of the point cloud registration.