Computer Engineering and Applications ›› 2020, Vol. 56 ›› Issue (7): 88-95.DOI: 10.3778/j.issn.1002-8331.1904-0174

Previous Articles     Next Articles

Point Cloud Registration Algorithm Based on Combination of NDT and ICP

WANG Qingshan, ZHANG Jun, LIU Yuansheng, ZHANG Xinchen   

  1. 1.Beijing Key Laboratory of Information Service Engineering, Beijing Union University, Beijing 100101, China
    2.College of Robotics, Beijing Union University, Beijing 100101, China
  • Online:2020-04-01 Published:2020-03-28

基于NDT与ICP结合的点云配准算法

王庆闪,张军,刘元盛,张鑫晨   

  1. 1.北京联合大学 北京市信息服务工程重点实验室,北京100101
    2.北京联合大学 机器人学院,北京100101

Abstract:

Lidar SLAM technology is an important way for unmanned vehicles to carry out precise navigation, and it is also a prerequisite to realize the safe driving of unmanned vehicles in complex unstructured road environment in park. A method of fast and accurate localization and mapping is constructed. Through a large amount of point cloud data returned by vehicle lidar, noise point removal and pre-processing of voxel grid filter are carried out to realize point cloud registration while maintaining the original point cloud shape. Firstly, the NDT(Normal Distribution Transform) point cloud registration algorithm is used to estimate the position and posture of the unmanned vehicle. Then, in order to estimate the unmanned vehicle position and posture precisely, it corrects the registered point cloud using the ICP(Iterative Closest Point) point cloud registration algorithm. Finally, on the basis of the above, the map update process is completed. This method only needs the vehicle-borne lidar sensor to realize the fast and high precision lidar SLAM. The algorithm is applied to small cyclone unmanned vehicle and verified in campus environment. The results show that the algorithm is reliable and effective.

Key words: Lidar SLAM, unmanned vehicle, Normal Distribution Transform(NDT), Iterative Closest Point(ICP), point cloud registration, localization, mapping

摘要:

Lidar SLAM技术是无人车进行精确导航的一种重要方式,也是实现无人车在复杂的园区非结构化道路环境中安全驾驶的一种前提保障。构建了一种快速精确定位与建图的方法,通过车载激光雷达返回的大量点云数据,进行噪声点移除以及Voxel Grid滤波的预处理,在保持原始点云形态的同时实现点云配准。首先利用NDT(Normal Distribution Transform)点云配准算法对无人车的位姿粗估计,然后利用ICP(Iterative Closest Point)点云配准算法对已配准的点云进行校正,实现无人车位姿的精确估计,进而完成地图的更新过程。该方法只需要车载激光雷达传感器实现了快速的、精度高的Lidar SLAM。将算法用于小旋风无人车,在校园环境进行了验证,结果表明该算法是可靠、有效的。

关键词: Lidar SLAM, 无人车, 正态分布变换(NDT), 迭代最近点(ICP), 点云配准, 定位, 地图构建