Computer Engineering and Applications ›› 2009, Vol. 45 ›› Issue (5): 164-166.DOI: 10.3778/j.issn.1002-8331.2009.05.048

• 图形、图像、模式识别 • Previous Articles     Next Articles

Comparison of several algorithms for tracking multiple dim point targets

WANG Bao-zhu,ASKAR Hamdulla   

  1. College of Information Science and Engineering,Xinjiang University,Urumqi 830046,China
  • Received:2008-08-13 Revised:2008-11-03 Online:2009-02-11 Published:2009-02-11
  • Contact: WANG Bao-zhu

几种典型的微弱点状多运动目标跟踪算法对比研究

王保柱,艾斯卡尔·艾木都拉   

  1. 新疆大学 信息科学与工程学院,乌鲁木齐 830046
  • 通讯作者: 王保柱

Abstract: The article has specially studied several typical algorithms for tracking multiple dim point targets in image sequences,although they can complete all-the-way tracking under different environment.Tracking performance exists big difference.There is high real-time tracking performance in PDA algorithm,but it can’t solve displace and coalescence in multi-target tracking;JPDA has solved the multi-target data association problem theoretically,but it is not implemented in reality because of big error of tracking and very high computational load;The algorithm based on maximum entropy fuzzy clustering amend the degrees of membership,the date association is high and efficiently avoid the errors of tracking and lost phenomenon.Through several typical algorithms’ simulation analysis,provide the reliable basis for the algorithm optimization.

摘要: 重点研究了序列图像情况下几种典型的微弱点状多运动目标实时跟踪算法,虽然它们都能够完成不同背景环境下目标的全程跟踪,但跟踪性能存在较大的差异,PDA算法具有较高的实时性,但容易出现目标的偏移和聚合现象;JPDA算法理论上解决了多目标数据关联问题,但跟踪过程存在较大误差且由于计算量大难以在工程中应用;基于最大熵高斯聚类算法对模糊隶属度进行了修正,数据关联性高且有效避免了目标的误跟和丢失现象。通过对几种典型算法的仿真分析,为多目标跟踪算法的优化提供可靠依据。