基于雷达测距的车辆识别与跟踪方法  

A Vehicle Identification and Tracking Method Based on Radar Ranging

作　　者：周俊静[1] 段建民[1] 杨光祖[1] 

机构地区：[1]北京工业大学电子信息与控制工程学院,北京100124

出　　处：《汽车工程》2014年第11期1415-1420,1414,共7页Automotive Engineering

基　　金：北京市教委基金(05002011200701)资助

摘　　要：本文中提出了一种基于雷达测距的车辆识别与跟踪方法,以准确地识别道路上的车辆目标。首先对采集到的每帧数据采用自适应距离阈值的方法进行聚类分析。在聚类内部采用端点迭代拟合算法提取线段,在此基础上得出目标的特征向量,并据之识别车辆目标。然后通过计算特征向量的代价函数进行多目标关联,利用卡尔曼滤波器根据关联结果更新目标的状态。最后,以智能车辆BJUT-IV配备LMS291激光雷达在校园道路上对本方法进行验证测试,结果表明,提出的方法能够准确地对多个行驶车辆进行跟踪。A method of vehicle identification and tracking based on radar ranging is proposed in this paper to accurately identify vehicle objects on road. Firstly a clustering analysis is conducted on data collected with an a-daptive distance threshold, and iterative end point fitting （ IEPF） algorithm is adopted to extract line segments with-in a cluster. On this basis the feature vector of object is obtained, with which vehicle object is identified. Then by calculating the cost function of feature vector, multi-object association is performed, and based on the results of as-sociation the state of object is updated with Kalman filter. Finally a verification test of the method is performed on the road in campus by an intelligent vehicle BJUT-IV with laser radar LMS291 . The results show that the method proposed can accurately track multiple moving vehicles.

关 键 词：车辆识别 雷达测距 线段提取 数据关联 

分 类 号：U495[交通运输工程—交通运输规划与管理]