基于相机雷达融合的改进GM-PHD多目标跟踪算法  

Improved GM-PHD multi-target tracking algorithm based oncamera and radar fusion

作　　者：张晗 李森 白傑 ZHANG Han;LI Sen;BAI Jie(School of Automotive Studies,Tongji University,Shanghai 201804,China)

机构地区：[1]同济大学汽车学院,上海201804

出　　处：《传感器与微系统》2020年第5期117-121,共5页Transducer and Microsystem Technologies

基　　金：国家重点研发计划资助项目(2018YFB0105002)。

摘　　要：针对复杂交通场景下单传感器跟踪性能不佳以及目标检测概率未知问题,提出了一种基于相机雷达融合的高斯混合概率密度假设(GM-PHD)多目标跟踪改进算法。通过预关联将目标集合划分为相机雷达量测匹配目标、仅相机量测匹配目标、仅雷达量测匹配目标以及无匹配目标,并采用不同的置信度对目标状态进行更新,综合雷达的径向距离以及相机的方位角对目标进行更准确的定位估计。将相机量测作为先验条件,简化优化高斯分量剪枝合并过程。仿真实验表明:所提算法能够有效提高目标跟踪精度和鲁棒性。Aiming at the problem of degenerative performance of single-sensor tracking method under complicated environment and the unknown detection probability,an improved Gaussian mixture probability hypothesis density(GM-PHD)multi-target tracking algorithm based on camera and radar fusion is proposed.Through pre-association,the target set is divided into both camera and radar matched targets,only-camera-matched targets,only-radar-matched targets and non-matched targets.According to that different degrees of confidence are applied to update target states.Integrating the distance measured by radar and the azimuth angle measured by camera,the position estimation of target is more accurate.Taking camera measurements as a priori condition,the merging and pruning process is simplified and improved.Simulation results indicate that the proposed algorithm is more accurate and robust.

关 键 词：多目标跟踪 多传感器融合 高斯混合-概率假设密度滤波器 匹配划分 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]