基于多目标灰狼优化的PHD粒子滤波改进算法  

MOGWO-based improved PHD particle filter algorithm

作　　者：吴冕 袁俊泉[1] 郑岱堃[1] 陈阿磊[1] WU Mian;YUAN Junquan;ZHENG Daikun;CHEN Alei(Air Force Early Warning Academy,Wuhan 430019,China)

机构地区：[1]空军预警学院,武汉430019

出　　处：《空军预警学院学报》2020年第2期104-110,共7页Journal of Air Force Early Warning Academy

基　　金：装备预研领域基金项目(61404130212)。

摘　　要：针对传统PHD粒子滤波算法存在目标数目估计精度低且位置估计误差大的缺陷,提出一种结合多目标灰狼优化的PHD(MOGWO-PHD)粒子滤波算法.该算法在预测和更新过程之间加入多目标灰狼优化算法,利用最新的观测信息将预测后的粒子集进行重新优化分布,使粒子移动至目标存在的高似然概率区域,减轻重采样后易出现的粒子贫乏问题;然后使用基于密度聚类(DBSCAN)算法对粒子进行聚类并提取目标状态.仿真结果表明,在不同杂波密度下,MOGWO-PHD算法对目标数目和状态的估计精度均优于传统PHD粒子滤波算法.The traditional probability hypothesis density(PHD)particle filter algorithm has the defects of low accuracy of target number estimation and large error of position estimation.In view of this problem,the paper proposes a PHD particle filter algorithm combined with multi-objective gray wolf optimization(MOGWO-PHD)algorithm.This algorithm adds the MOGWO algorithm between the prediction and update process,and uses the latest observation information to re-optimize the distribution of the predicted particle sets,causing the particle to move to the high likelihood probability region where targets exist,so as to reduce the particle poverty after re-sampling.And then,the DBSCAN algorithm is used to cluster the particles and extract the target state.Simulation results show that under different clutter densities,MOGWO-PHD algorithm’s estimation accuracy of target number and state is better than that of traditional PHD particle filter algorithm.

关 键 词：概率假设密度 粒子滤波 多目标灰狼优化 密度聚类 

分 类 号：TN957[电子电信—信号与信息处理]