基于车载4线激光雷达的前方道路可行驶区域检测  

Preceding Drivable Area Detection Based on Four-layer Laser Radar

作　　者：段建民[1] 李龙杰[1] 郑凯华[1] 

机构地区：[1]北京工业大学北京市交通工程重点实验室,北京100124

出　　处：《汽车技术》2016年第2期55-62,共8页Automobile Technology

基　　金：北京市属高等学校人才强教计划资助项目(038000543115025)

摘　　要：通过分析大量激光雷达数据,提出了改进的DBSCAN算法对雷达数据进行聚类,以使智能汽车获得前方道路和障碍物信息。根据不同的密度参数,多次调用该算法完成多密度聚类。结合提出的道路边沿、路面和障碍物等信息提取方法,在结构化或半结构化的城市道路中对提出的方法进行实车试验,结果表明该算法可实时准确的提取智能汽车的可行驶区域信息。By analyzing a large number of laser radar＇s data, a modified Density Based Spatial Clustering of Applications with Noise（DBSCAN） algorithm is proposed to obtain the information on roads and obstacles for intelligent vehicle. According to different density parameters, this algorithm is called to complete multiple density clustering. Also,methods for extracting the information on road edge, road surface and obstacles etc., are proposed, which are tested on vehicle on structured and semi- structured urban roads. The results show that the methods can real- time extract the accurate information on drivable area for intelligent vehicle.

关 键 词：4线激光雷达 DBSCAN算法 加权欧式距离 可行驶区域 

分 类 号：U463.6[机械工程—车辆工程]