联合模糊聚类和拟蒙特卡罗重采样的群目标跟踪算法

Group targets tracking algorithm by combination of fuzzy clustering and Quasi-Monte Carlo resampling method

中文关键词  群目标  跟踪  滤波  模糊聚类  拟蒙特卡罗  重采样  数据关联

英文关键词  group targets  tracking  filter  fuzzy clustering  Quasi-Monte Carlo  resampling  data association

中文摘要

      联合概率数据关联粒子滤波(joint probabilistic data association-particle filter, JPDA--PF)算法常被用来解决群目标跟踪中的数据关联和非线性滤波问题. 针对算法的数据关联时间较长以及样本枯竭问题, 本文阐述了一种利用模糊聚类和拟蒙特卡罗重采样的群目标跟踪算法. 首先, 在群演化网络模型的基础上, 采用最大熵模糊聚类法来完成群内个体目标和量测之间的数据关联, 利用模糊隶属度来构建互联概率矩阵. 其次, 在目标状态样本的重采样的过程中, 利用随机化拟蒙特卡罗序列映射到拟复制样本的子空间上, 提高样本的多样性, 抑制样本枯竭的出现. 仿真实验结果表明, 与JPDA--PF算法相比, 本文算法能有效估计群内目标状态和群结构, 并具有更优的估计性能.

英文摘要

      Joint probabilistic data association-particle filter (JPDA–PF) algorithm was always used to solve the data association and nonlinear filtering problem. Aiming at the high computational complexity in data association and the sample impoverishment problem in resampling step, an improved algorithm by combination of fuzzy clustering and Quasi-Monte Carlo resampling method was proposed for group targets tracking in this paper. First, based on the group evolving network model, the maximum entropy fuzzy clustering was used to achieve the data association between group individual targets and measurements, and the association probability matrix was constructed by the fuzzy membership degree. Then, the randomized Quasi-Monte Carlo points were transformed into some independent sub-spaces of planned duplicate particles to improve the diversity of samples and prevent the occurrence of sample impoverishment. The computer simulations showed that compared the JPDA–PF algorithm, our proposed algorithm can estimate group targets state and group structure effectively, and obtain the better estimation performance.