结合匈牙利指派和改进粒子滤波的多目标跟踪算法

使用汽车雷达进行多目标跟踪时，为了提高航迹关联效率并改善非线性场景跟踪效果，提出了结合匈牙利指派和卡尔曼重要性采样的粒子滤波(Particle Filter with Kalman Importance Sampling，PF-KIS)算法。首先，将航迹关联分解为聚类和指派，通过密度聚类筛选并整合有效目标，经过匈牙利指派得到目标和航迹的最佳匹配关系，避免产生多余联合事件，提高关联效率；其次，以卡尔曼滤波的结果作为粒子滤波的先验，使采样粒子分布更合理，提高估计精度，进而改善非线性跟踪能力。实验表明，算法平均航迹关联正确率约为95%；非线性场景误差约为卡尔曼滤波的1/2，有效地改善了非线性场景跟踪能力。

A multi-target tracking algorithm combining Hungarian assignment and improved particle filter

When using vehicle radar for multi-target tracking,in order to improve the track correlation efficiency and tracking performance in nonlinear scenes,an algorithm based on particle filter with Kalman importance sampling(PF-KIS) and Hungarian assignment is proposed.Firstly,track association is divided into clustering and assignment.The effective target is screened and integrated by density clustering.And the best matching relationship between target and track is obtained through Hungary assignment,which avoids generating redundant joint events so as to improve the association efficiency.Secondly,the result of Kalman filter is used as the prior of particle filter,which makes the sampling particle distribution more reasonable,improves the estimation precision,then proceeds to the next step to improve the nonlinear tracking ability.Experiments show that the average track correlation precision of the proposed algorithm is about 95%;the nonlinear scene error is about 1/2 of the Kalman filter,which effectively improves the nonlinear scene tracking ability.