航空学报 > 2014, Vol. 35 Issue (4): 1091-1101   doi: 10.7527/S1000-6893.2013.0439

区域杂波估计的多目标跟踪方法

瑚成祥, 刘贵喜, 董亮, 王明, 张菁超   

  1. 西安电子科技大学 机电工程学院, 陕西 西安 710071
  • 收稿日期:2013-05-30 修回日期:2013-10-30 出版日期:2014-04-25 发布日期:2013-11-06
  • 通讯作者: 刘贵喜,Tel.:13700296049 E-mail:gxliu@xidian.edu.cn E-mail:gxliu@xidian.edu.cn
  • 作者简介:瑚成祥男,硕士研究生。主要研究方向:多目标跟踪,跟踪滤波。 E-mail:huchengxiang013@163.com;刘贵喜男,博士,教授,博士生导师。主要研究方向:信息融合,目标跟踪,计算机视觉。Tel:13700296049 E-mail:gxliu@xidian.edu.cn
  • 基金资助:

    国家级项目(9140A******13DZ01)

Region Clutter Estimation Method for Multi-target Tracking

HU Chengxiang, LIU Guixi, DONG Liang, WANG Ming, ZHANG Jingchao   

  1. School of Mechano-electronic Engineering, Xidian University, Xi'an 710071, China
  • Received:2013-05-30 Revised:2013-10-30 Online:2014-04-25 Published:2013-11-06
  • Supported by:

    National Level project (9140A******13DZ01)

摘要:

高斯粒子概率假设密度(PHD)滤波往往假定杂波密度参数已知,这种做法对于实际应用是不现实的。此外,杂波的参数值通常依赖于环境条件,可能随时间发生变化。因此,多目标跟踪算法中需要实时准确估计杂波密度的参数。基于此,提出了一种多目标跟踪的区域杂波估计方法。首先根据量测信息在线估计出场景中的杂波数目,然后估计落入目标附近感兴趣区域的杂波数,并估计每个目标感兴趣区域杂波强度。仿真结果表明,在复杂场景下算法的跟踪性能明显优于未进行杂波估计的多目标跟踪算法,提高了跟踪的实时性和跟踪精度。

关键词: 概率假设密度, 目标跟踪, 粒子滤波, 杂波估计, 随机有限集

Abstract:

Gaussian mixture particle probability hypothesis density (PHD) filter often assumes that the clutter density parameters are known. This method is impractical for real applications. In addition, the parameter values of the clutter points are usually dependent on environmental conditions, and they may change over time. Therefore, it is desirable for multiple-target tracking algorithm in real time to estimate the clutter density parameters. In this paper, a method of the clutter estimation about multi-target tracking is presented. Firstly, we estimate the number of clutter points in the scene online. Secondly, we estimate the clutter number and intensity in each region of interest. Simulation results show that its tracking performance is much better than those of multiple-target tracking algorithms which have not estimated the clutter intensity in complex situations and that it improves the real-time tracking and tracking accuracy.

Key words: probability hypothesis density, target tracking, particle filter, clutter estimation, random finite set

中图分类号: