边缘化粒子概率假设密度滤波的多目标跟踪  

Multi-target Tracking Based on Marginal Particle Filtering-Probability Hypothesis Density

作　　者：于洋[1] 宋建辉[1] 刘砚菊[1] 司冠楠 

机构地区：[1]沈阳理工大学自动化与电气工程学院,沈阳110159

出　　处：《火力与指挥控制》2017年第5期14-18,22,共6页Fire Control & Command Control

基　　金：国家自然科学基金(61373089);辽宁省教育厅基金资助项目(LT2012005)

摘　　要：针对复杂情况下的多目标跟踪问题,提出一种边缘化粒子概率假设密度滤波(MPF-PHD)方法。该方法首先将复杂情况下多个目标的状态向量分别提取出其中的非线性状态与线性状态。然后利用粒子概率假设密度滤波(PF-PHD)估计非线性状态,利用卡尔曼滤波(KF)估计线性状态,并把其中与非线性状态相关的线性状态估计用来优化非线性状态估计。通过对MPF-PHD方法与传统的PF-PHD方法仿真对比,验证了MPF-PHD方法有效解决了复杂情况下多目标跟踪的漏检问题,提高了多目标状态估计精度。In view of the multi-target tracking problem under complex conditions, this paper puts forward a kind of marginalized particle filtering-probability hypothesis density （MPF-PHD） method. Firstly the nonlinear and linear state in the state vector of multi-target under the complex conditions is extracted separately in this method. Then using particle probability hypothesis density filter （PF-PHD） nonlinear state estimation, Kalman Filter （KF） is used to estimate the linear state, and the linear state estimation of nonlinear state related is adopted to optimize the nonlinear state estimation. According to contrast MPF-PHD method with the traditional PF-PHD method simulation, it is verified that the MPF- PHD method can effectively solve the multi-target tracking under complex conditions of residual problem, the multiple target state estimation accuracy is improved.

关 键 词：边缘化粒子概率假设密度滤波 多目标跟踪 非线性状态估计 卡尔曼滤波 

分 类 号：TN953[电子电信—信号与信息处理]