摘要
针对粒子概率假设密度滤波(P-PHDF)算法估计精度低、滤波发散和粒子退化的缺陷,提出了一种无迹粒子PHD滤波(UP-PHDF)算法。该算法以UKF算法产生重要性函数并从中采样,通过观测值更新粒子的权值,再用加权的粒子估计PHD函数,进而得到优化的状态估计均值和方差进行传播。最后,对UP-PHDF算法进行了分析和实现,并将该算法和P-PHDF算法进行了比较。仿真结果表明,UP-PHDF算法不仅大大提高了滤波估计的精度,同时提高了跟踪系统的稳定性和鲁棒性。
In order to avoid the weakness such as low accurate, filter divergence and particle degradation of particle probability hypothesis density filter (P-PHDF) algorithm, an unscented particle PHI) filter (UP- PHDF) algorithm was proposed. Firstly, the importance density function was generated and sampled by the Unscented Kalman filter (UKF) method. Then, the weighted particles was adopt to approximate PHD function and further estimated the optimized state value and covariance, which preferably resolved the problem of multi- target tracking under nonlinear and clutter environment. Finally, in the complex environment, the UP-PHDF algorithm was analyzed and realized. In addition, the proposed algorithm was compared with P-PHDF in multi-target tracking simulation experiment. The results show that, the filtering estimate accuracy of the proposed algorithm is significantly improved, as well as the stability and robustness of the tracking system.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2013年第1期94-98,103,共6页
Journal of System Simulation
基金
国家科技部支撑课题(2011AA110201)
关键词
多目标跟踪
随机有限集
无迹粒子滤波
粒子概率假设密度滤波
multi-target tracking
random finite sets
tmscented particle filter
particle probability hypothesisdensity filter