引用本文
  • 裴家正,黄勇,董云龙,何友,陈小龙,陈唯实. 基于PHD的粒子滤波检测前跟踪改进算法[J]. 雷达科学与技术, 2019, 17(3): 263-270.    [点击复制]
  • PEI Jiazheng,HUANG Yong,DONG Yunlong,HE You,CHEN Xiaolong,CHEN Weishi. PHD-Based Particle Swarm Optimization Particle Filter Radar Track-Before-Detect Algorithm[J]. Radar Science and Technology, 2019, 17(3): 263-270.   [点击复制]
【打印本页】 【HTML】 【下载PDF全文】 查看/发表评论 下载PDF阅读器关闭

←前一篇|后一篇→

过刊浏览    高级检索

本文已被:浏览 891次   下载 154 本文二维码信息
码上扫一扫!
基于PHD的粒子滤波检测前跟踪改进算法
裴家正,黄勇,董云龙,何友,陈小龙,陈唯实
0
(1.海军航空大学, 山东烟台 264001;2.中国民航科学技术研究院机场研究所, 北京 100028)
摘要:
针对在低信噪比目标检测问题中,基于PHD的粒子滤波检测前跟踪算法(PHD-TBD)存在目标位置估计误差较大的缺陷,提出一种结合粒子群优化算法的基于PHD的粒子滤波检测前跟踪方法(PSO-PHD-TBD)。该算法在滤波预测和更新步骤之间加入基于NSGA-Ⅱ的多目标粒子群优化算法,结合量测信息将预测完成的粒子集的分布进行优化,将所有粒子转移到后验概率密度较大的区域,进而改善了多目标位置估计的性能;然后使用基于密度聚类的DBSCAN算法对粒子聚类,提取目标状态。仿真实验表明,在不同信噪比条件下,PSO-PHD-TBD在多目标数目估计情况与PHD-TBD算法一致,而位置估计精度明显优于PHD-TBD算法。
关键词:  概率假设密度  粒子滤波  粒子群优化  基于密度聚类  检测前跟踪
DOI:DOI:10.3969/j.issn.1672-2337.2019.03.005
基金项目:国家自然科学基金(No.U1633122,61501487,61871392,61471382,61531020);国防科技基金(No.2102024);中国博士后科学基金(No.2017M620862);“泰山学者”和中国科协“青年人才托举工程”专项经费资助(No.YESS20160115)
PHD-Based Particle Swarm Optimization Particle Filter Radar Track-Before-Detect Algorithm
PEI Jiazheng,HUANG Yong,DONG Yunlong,HE You,CHEN Xiaolong,CHEN Weishi
(1. Naval Aviation University, Yantai 264001, China; 2. Airport Research Institute, China Academy of Civil Aviation Science and Technology, Beijing 100028, China)
Abstract:
In the target detection under low signal-to-noise ratio, the PHD-based particle filter track-before-detect (PHD-TBD) algorithm has large estimation error of target position. A PF-PHD-TBD algorithm combined with particle swarm optimization is proposed. The algorithm adds a multi-objective particle swarm optimization algorithm based on NSGA-II between the filtering prediction and the updating steps. The measurement information is used to optimize the distribution of the predicted particle sets, and all of the particles are transferred to the high-posterior probability density to improve the performance of multi-target location estimation. Then the DBSCAN algorithm based on density clustering is applied to cluster the particles and extracts the target state simulation. The experiments under different SNR conditions show that the PSO-PHD-TBD is consistent with the PHD-TBD algorithm in terms of the multi-target number estimation, and its position estimation accuracy is significantly higher than the PHD-TBD algorithm.
Key words:  probability hypothesis density  particle filter  particle swarm optimization  density based spatial clustering of applications with noise (DBSCAN)  track-before-detect