基于数值积分粒子PHD多目标跟踪滤波算法  

PHD Filter Algorithm for Multi-target Tracking Based on Quadrature Particle 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,西安710077

出　　处：《电光与控制》2015年第11期23-26,41,共5页Electronics Optics & Control

基　　金：陕西省自然科学基金(2011JM.8023)

摘　　要：在多目标跟踪环境下,粒子概率假设密度(P-PHD)因杂波、漏检和非线性情况,导致滤波精度不高,跟踪发散,对此问题提出一种基于数值积分粒子概率假设密度滤波算法。利用数值粒子滤波(QPF)实现概率假设密度(PHD),用数值卡尔曼滤波(QKF)算法得到更好的重要性密度函数,并从中采样得到粒子,使粒子的分布更接近真实的概率假设密度分布。试验仿真表明,与粒子概率假设密度和容积粒子概率假设密度滤波算法相比,所提算法的滤波精度和稳定性明显提高。Under the circumstances of multi-target tracking, due to clutter, missed detection and no- linearity, the Particle Probability Hypothesis Density （P-PHD） algorithm will result in such problems as low filter accuracy and tracking divergence. To overcome these problems, aQuadrature Particle based PHD Filter （QPF-PHD） algorithm is proposed. The Quadrature Particle Filter （QPF） is used to realize PHD; the Quadrature Kalman Filter （QKF） algorithm is applied to generate a better importance density function, from which to sample particles, enabling the distribution of particles to be much closer to the real distribution of PHD. The simulation results demonstrate that：compared with particle PHD and Gizmo particle PHD filter algorithms, the proposed filter algorithm has higher filter accuracy and stability.

关 键 词：多目标跟踪 数值积分粒子滤波 概率假设密度滤波 随机有限集 

分 类 号：TN953[电子电信—信号与信息处理]