作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2018, Vol. 44 ›› Issue (7): 279-284. doi: 10.19678/j.issn.1000-3428.0047524

• 开发研究与工程应用 • 上一篇    下一篇

基于边缘卡尔曼滤波的GM-PHD多目标被动跟踪算法

曲长文 1,冯奇 1,毛宇 2,周强 1   

  1. 1.海军航空工程学院 电子信息工程系,山东 烟台 264001; 2.中国人民解放军 91431部队,海南 文昌 571300
  • 收稿日期:2017-06-08 出版日期:2018-07-15 发布日期:2018-07-15
  • 作者简介:曲长文(1963—),男,教授、博士生导师,主研方向为无源定位与跟踪技术、合成孔径雷达成像、目标检测与识别;冯奇,博士研究生;毛宇,学士;周强,助理研究员、博士。

GM-PHD Multi-objective Passive Tracking Algorithm Based on Margin Kalman Filtering

QU Changwen  1,FENG Qi  1,MAO Yu  2,ZHOU Qiang  1   

  1. 1.Department of Electronic and Information Engineering,Naval Aeronautical and Astronautical University,Yantai,Shandong 264001,China; 2.91431 Troops of PLA,Wenchang,Hainan 571300,China
  • Received:2017-06-08 Online:2018-07-15 Published:2018-07-15

摘要:

针对杂波干扰条件下,非线性、个数时变的多目标被动跟踪问题,提出一种基于边缘卡尔曼滤波的高斯混合概率假设密度(PHD)滤波算法。采用边缘化变换计算目标状态的概率分布特性,获得目标状态及其协方差矩阵估计的闭式解,解决目标模型非线性问题。利用量测信息生成新生目标强度,使滤波器具备对观测空间任意位置随机出现新目标的跟踪能力。实验结果表明,与扩展卡尔曼PHD算法、无迹卡尔曼PHD算法和容积卡尔曼PHD算法相比,该算法在生成目标轨迹、目标个数估计和跟踪精度等方面有更好的性能。

关键词: 多目标跟踪, 随机有限集, 边缘卡尔曼滤波, 概率假设密度, 量测驱动

Abstract:

For the problem of passive multi-objective tracking with the time varying number in clutter,a Gaussian mixture Probability Hypothesis Density(GM-PHD) based on the marginalized Kalman filter for nonlinear Gaussian system is proposed.The marginalized transformation is applied to calculate the prediction and update distribution of target states for the nonlinear multi-objective models,and then the close-form solution of the target state and its covariance can be got.The measurements are used to generate the density of new born targets that appear randomly anywhere in the observation space.Experimental results show that compared with the Extended Kalman PHD(EK-PHD),Unscented Kalman PHD(UK-PHD) and Centralized Kalman PHD(CK-PHD),the margin Kalman GM-PHD(MK-PHD) is better in aspects of generating target trajectory,estimation of target number and tracking precision.

Key words: multi-objective tracking, Random Finite Sets(RFS), margin Kalman filtering, Probability Hypothesis Density(PHD), measurement driving

中图分类号: