摘要
针对真实交通场景中障碍物的检测与跟踪问题,提出了一种基于三维激光雷达HDL-64E的无人车障碍物检测与跟踪方法。首先对三维激光雷达产生且经路面分割后的点云数据栅格化,并进行栅格增补。在障碍物聚类之后,先利用无人车RTK-GPS数据和INS航向角数据进行多帧融合的静态障碍物的检测,再进一步利用动态障碍物模板匹配算法对静态障碍物检测结果形成的可行驶区域进行动态障碍物检测。最后,利用标准卡尔曼滤波器实现对动态障碍物的跟踪。本文方法应用在自主研发的无人车上的大量实验结果表明,本文提出的方法具备较高的可靠度,满足无人车的环境感知要求。
Aiming at the problem of detection and tracking of obstacles in real traffic scenarios,a method of detecting and tracking unmanned vehicles based on 3D laser radar HDL-64E is proposed.Firstly,the point cloud data generated by 3D laser radar is rasterized after road segmentation,with the raster supplemented.After obstacle clustering,the detection of static obstacles is conducted with multi-frame fusion based on real-time kinematic GPS data and the azimuth angle data in inertial navigation system.Furthermore,dynamic obstacle detection is performed by using dynamic obstacle template matching algorithm on travelable areas determined by static obstacle detection.Finally,the standard Kalman filter is adopted to track dynamic obstacles.The results of a large number of experiments in applying the method to a self-developed unmanned vehicle show that the method proposed has a high reliability,meeting the environmental perception requirements for unmanned vehicles.
作者
谢德胜
徐友春
王任栋
苏致远
Xie Desheng;Xu Youchun;Wang Rendong;Su Zhiyuan(Postgraduate Training Brigade,Army Military Transportation University,Tianjin 300161;Department of Military Vehicle,Army Military Transportation University,Tianjin 300161)
出处
《汽车工程》
EI
CSCD
北大核心
2018年第8期952-959,共8页
Automotive Engineering
基金
国家自然科学基金重大项目(91220301)
国家973计划项目(2016YFB0100903)资助
关键词
无人车
三维激光雷达
障碍物检测
动态障碍物跟踪
unmanned vehicles
3D laser radar
obstacle detection
dynamic obstacle tracking