摘要
针对无人车驾驶车辆使用相机进行环境感知时容易受到光照、阴影等影响,导致难以获得有效的障碍物种类及位置信息等问题,使用64线激光雷达获取前方6~60 m,左、右9 m范围内的点云数据,使用k-d树的方法将映射到图像的三维激光点云进行排列并插值,对激光雷达数据与图像信息进行融合。为解决映射后图像深度信息稀疏的问题,采用二维图像的纹理信息丰富雷达点云的方法对稀疏的深度图像进行拟合插值,得到分辨率为1 624×350的小粒度深度图像,为无人车行驶提供准确的环境场景信息。
Since the unmanned vehicle is easily influenced by illumination and shadow while perceiving environment with camera,and it is difficult to acquire effective obstacles variety and location,the paper firstly acquires point cloud data in the range of 6 ～ 30 m ahead and 9 m at left and right with HDL-64 E. Then,it arranges and interpolates three-dimensional laser-point cloud which maps to image with k-d tree,and integrates LIDAR data with image information. Finally,in order to solve the sparse depth information of the image after mapping,it obtains the range image of 1 624 × 350 by making fitting and interpolation on the sparse range image with the method of enriching radar point cloud by the texture information of twodimensional image,which can provide accurate information of environment scenario for unmanned vehicle.
出处
《军事交通学院学报》
2017年第10期80-84,共5页
Journal of Military Transportation University
基金
国家重大研发计划"智能电动汽车路径规划及自主决策方法"(2016YFB0100903)
国家自然科学基金重大项目"自主驾驶车辆关键技术与集成验证平台"(91220301)
关键词
无人车
视觉
激光点云
深度图像
unmanned vehicle
vision
laser point cloud
range image
作者简介
王东敏（1992-）,男,硕士研究生;
彭永胜（1972-）,男,博士,教授,硕士研究生导师.
