基于三维激光雷达和深度图像的自动驾驶汽车障碍物检测方法  

-Obstacle detection based on 3D laser scanner and range image for intelligent vehicle

作　　者：王新竹[1] 李骏[1,2] 李红建[2] 尚秉旭[2] 

机构地区：[1]吉林大学汽车工程学院,长春130022 [2]中国第一汽车集团公司技术中心,长春130021

出　　处：《吉林大学学报（工学版）》2016年第2期360-365,共6页Journal of Jilin University:Engineering and Technology Edition

基　　金：'973'国家重点基础研究发展计划项目(2012CB723801)

摘　　要：介绍了一种基于三维激光雷达和深度图像的障碍物检测方法。首先,根据Velodyne HDL-32E激光雷达自身工作特性,将点云数据以矩阵方式表达,并表示为深度图像;然后,根据点云中各点的距离信息在深度图像横向上进行聚类;最后,在纵向上建立线性模型,对聚类点进行分类,划分出地面点集和障碍物点集。仿真实验结果表明:本方法能够抑制障碍物遮挡造成的误判,并能够很好地适应地形变化。An obstacle detection method based on 3Dlaser scanner and range image is proposed for intelligent vehicle.First,the range image of one scan is established based on the Welodyne HDL-32 E laser scanner data.Then,according to the range image analysis,the point cloud in one row of the image is classified into several groups.Finally,the point clouds in different groups are detected by a linear model in column.Simulation reveals that using the proposed method,the rate of false detection caused by the obstacle lock is reduced,and this method well adapts to the terrain change.

关 键 词：车辆工程 三维点云 障碍物检测 环境感知 

分 类 号：U461[机械工程—车辆工程]