基于迭代积分卡尔曼PHD滤波多目标跟踪  

Probability Hypothesis Density Filter Algorithm Based on Iterated Quadrature Kalman Filter

作　　者：穆祥强 王朝英[1] 危璋 孔云波[1] 

机构地区：[1]空军工程大学信息与导航学院,陕西西安710011

出　　处：《现代防御技术》2016年第4期95-100,共6页Modern Defence Technology

基　　金：陕西省自然科学基金(2011JM.8023)

摘　　要：在多目标跟踪环境下,通过对数值积分卡尔曼概率假设密度滤波器(QK-PHD)的误差分析,提出一种迭代QK-PHD算法。该算法基本思想是将一步预测作为第1次迭代的初始值,通过量测更新得到状态估计,然后将此状态估计作为下一步迭代的初始均值和方差。通过多次迭代后滤波输出精度更高方差更小。Matlab仿真结果表明,在不明显增加计算时间的基础上,迭代QK-PHD比QK-PHD的跟踪效果更加明显,滤波输出精度更高。Under the condition of multi-target tracking, by analyzing the error of quadrature Kalman probability hypothesis density filter （ QK-PHD）, the iterated QKF algorithm is proposed. The IQKF-PHD is used to re-estimate the one-step state and measurement predictions by virtue of the measurement upda- ted state variable, and then the mean and variance of state variable are estimated. After several itera- tions, the outputs of IQKF-PHD have higher accuracy and less variance. Matlab simulation results dem- onstrate that the tracking accuracy and the outputs of IQKF-PHD are superior to those of QKF-PHD.

关 键 词：迭代 多目标跟踪 状态估计 数值积分 卡尔曼滤波 概率假设密度滤波 

分 类 号：TN713[电子电信—电路与系统] TP391.9[自动化与计算机技术—计算机应用技术]