3D Point Cloud Registration for Multiple Roadside LiDARs with Retroreflective Reference

In intelligent transportation systems, LiDAR has been used to acquire traffic information on the roadside. Due to the sensing range and occlusions between vehicles, single LiDAR can only be applied in simple scenes and limited scope. In this paper, multiple LiDARs are applied to solve the problems of traffic information sensing in the complex traffic environment. A new point cloud registration method is proposed. This method combines the advantages of the iterative closest point (ICP) algorithm and the Zhang's calibration method for camera calibration. First of all, a reference system is made for registration, so that the registration of two sets of points is converted to the registration of reference points with different coordinates. Second, filtering based on intensity is conducted to extract the points on the reference system. To remove noises, we apply the density-based spatial clustering of applications with noise (DBSCAN) algorithm for denoising in this paper. Then, a robust ICP algorithm based on M-estimation is applied to realize the registration of reference points in two coordinate systems. Finally, this method has been demonstrated by some experiments in real traffic scenes, experiment results show that the proposed method can achieve accurate registration of point cloud data from multiple LiDARs. Besides, the convergence time of this method is about 10 seconds, which can achieve better performance compared with traditional point registration methods.