基于多传感融合的目标跟踪方法研究  

Research on Target Tracking Method Based on Multi-sensor Fusion

作　　者：刘志强[1] 张光林 邱惠敏 LIU Zhiqiang;ZHANG Guanglin;QIU Huimin(School of Automotive and Traffic Engineering,Jiangsu University,Zhenjiang 212013,China;Traffic Management Science Research Institute of the Ministry of Public Security,Wuxi 214106,China)

机构地区：[1]江苏大学汽车与交通工程学院,江苏镇江212013 [2]公安部交通管理科学研究所,江苏无锡214106

出　　处：《重庆理工大学学报（自然科学）》2021年第2期1-7,共7页Journal of Chongqing University of Technology：Natural Science

基　　金：江苏省研究生实践创新计划项目(SJCX19_1168)。

摘　　要：针对单一传感器对前方车辆识别准确率低的问题,基于多传感器融合模型建立了1种目标车辆识别方法。首先开展摄像头与毫米波雷达的联合标定,实现多传感器在时空上的融合;然后引入全局最近邻算法对雷达和摄像头各自采集的目标序列与跟踪目标进行数据匹配,确定跟踪目标的2组局部估计;最后通过D-S证据理论对2组目标序列进行优化组合,获取车辆行驶状态的最优结果,从而实现对目标的识别。通过Matlab/Simulink联合搭建试验平台对所研究的融合模型进行算法验证。试验结果表明:该融合算法在不同天气条件下对目标的平均检测率为88.3%,可实现对目标车辆的准确识别与跟踪。Aiming at the problem of low accuracy of a single sensor in identifying the vehicle ahead,a target vehicle identification method is established based on a multi-sensor fusion model.Firstly,the joint calibration of millimeter-wave radar and camera is carried out to realize the fusion of multi-sensor in time and space.Then,the global nearest neighbor(GNN)algorithm is introduced to match the target sequence collected by radar and camera with the tracking target,and two sets of target sequences are determined.Finally,D-S evidence theory is established to fuse the local estimation values to obtain the optimal estimation value of the target,so as to achieve the target recognition.The proposed fusion model is validated by building a vehicle tracking verification platform with Matlab/Simulink.The experimental results show that the average detection rate of the fusion algorithm is 88.3%under different weather conditions,which can realize the accurate recognition and tracking of the target vehicle.

关 键 词：信息融合 数据匹配 全局最近邻 D-S证据理论 目标识别 

分 类 号：U461[机械工程—车辆工程]