Point cloud registration with 2D and 3D fusion information on mobile robot integrated vision system

Abstract

Multi-view point clouds registration method based on the 2D and 3D fusion information is proposed. First the joint calibration of the CCD camera and 3D laser radar system is conducted to fuse the image information and range information. Then, matching feature points in the images are obtained with the SIFT algorithm. Based on the joint calibration results and the images feature matching points, the point clouds feature matching points are determined. After that, the initial value of the different view point cloud space transform is calculated with the RANSAC, and ICP algorithm is adopted to optimize the results. Finally, the proposed method was implemented on a mobile robot and verified effective in the experiment.