Real-Time Dead Reckoning and Mapping Approach Based on Three-Dimensional Point Cloud

摘要

Light detection and ranging equipment (LiDAR) has wide applications in the field of mobile surveying, autonomous driving, unmanned aerial vehicle and military, etc., since its abilities of fast three-dimensional environmental information acquisition, robustness to variable illumination conditions as well as wide measurement range. The main difficulty for localization and mapping based on LiDAR is the registration between successive point clouds, which is caused by the continuous motion and undetermined trajectory of LiDAR. A real-time dead reckoning and three-dimensional environment mapping method based on 3D-LiDAR is presented in this paper. The proposed system adopts a divide and parallel method, which performs high frequency pose estimation and low frequency mapping on parallel threads to ensure the real-time performance. In point preprocess section, the efficient elimination of point cloud distortion is processed and feature points are extracted. In dead reckoning section, point cloud registration employs generalized iterative closest point (GICP) algorithm which increases accuracy of registration by taking local covariance information of each point into consideration. In environmental mapping section, the multi-channel GICP (MCGICP) algorithm which adds point cloud intensity information into GICP framework is applied to align local point cloud with the global point cloud map. The robustness and convergence are promoted, since the fusion of additional point in-formation. Finally, the performance of the proposed approach is evaluated by the KITTI dataset. The experimental results show that our proposed simultaneous dead reckoning and 3D environment mapping method based on 3D laser scanner is an effective and feasible on-line solution which achieves high accuracy under 1% in all kinds of environment scenarios. Moreover, the ability of self-localization and real-time mapping is improved dramatically in those environments with poor geometric feature patterns.