Computer Engineering and Applications ›› 2009, Vol. 45 ›› Issue (8): 158-160.DOI: 10.3778/j.issn.1002-8331.2009.08.048

• 数据库、信号与信息处理 • Previous Articles     Next Articles

Algorithm based on fuzzy clustering and particle filter for tracking multiple cross-moving point targets

Askar,YU Wei-jun,WANG Xin-bin,LIU Deng-feng   

  1. College of Information Science and Engineering,Xinjiang University,Urumqi 830046,China
  • Received:2008-01-24 Revised:2008-05-26 Online:2009-03-11 Published:2009-03-11
  • Contact: Askar

模糊聚类粒子滤波的点状交叉多目标跟踪算法

艾斯卡尔,于伟俊,王新滨,刘登峰   

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

Abstract: Present a new low Signal to Noise Ratio(SNR) infrared image sequences multitarget detection and tracking algorithm,which combines TBD detection algorithms and fuzzy clustering particle filter tracking algorithm.First through multi-frame TBD process,detect initial position and velocity of moving objects,and then estimate target state during tracking stages of fuzzy clustering particle filter,and open a window for testing,probability integration.Infrared images of real sequence simulation,simulation results show that the algorithm has good real-time and high accuracy.

Key words: Thack-Before-Detect(TBD), particle filter, maximum fuzzy entropy Gaussian clustering, multitarget, data fusion

摘要: 提出了一种新的低信噪比红外序列图像多目标检测跟踪算法,该算法有机地结合了TBD检测算法与模糊聚类粒子滤波跟踪算法。首先通过多帧TBD处理后,检测出运动目标的初始位置、运动速度,然后在跟踪阶段采用粒子滤波算法估计目标运动状态,并在估计位置开一个跟踪窗进行检测、模糊聚类概率融合。对真实红外图像序列进行实验仿真,仿真结果验证了该算法具有良好的实时性与很高的精确性。

关键词: 检测前跟踪, 粒子滤波器, 最大模糊熵高斯聚类, 多目标, 数据融合