计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (2): 64-64.

• 学术探讨 • 上一篇    下一篇

联合轮廓法在低分辨率视频下的多目标追踪

程光,罗予频,王宏宝   

  1. 清华大学自动化系
  • 收稿日期:2006-01-24 修回日期:1900-01-01 出版日期:2007-01-11 发布日期:2007-01-11
  • 通讯作者: 程光 seagle seagle

Multiple Objects Tracking in Low Resolution Based on Joint Contour Method

,Yupin Luo,Hongbao Wang   

  1. 清华大学自动化系
  • Received:2006-01-24 Revised:1900-01-01 Online:2007-01-11 Published:2007-01-11

摘要: 相似目标间的遮挡一直是多目标追踪问题中的难点。针对这个问题本文提出了联合轮廓法。这种方法在使用粒子滤波器的基础上,通过对遮挡目标集的外轮廓建模来确定处于遮挡关系中目标的状态,从而近一步得到目标的轨迹。算法在实验中显示了良好的鲁棒性,并且计算复杂度与遮挡目标的个数成二次多项式关系。

关键词: 多目标追踪, 粒子滤波器, 联合轮廓法, 主动形状模型

Abstract: Tracking multiple similar objects is a challenging task, expecially in resolution, because of occlusion between objects. The Joint Contour Method (JCM) based on the particle filter is proposed to solve the problem of partial occlusion in this paper. It is a method to determine the object status by modeling the contour of the set of objects in occlusion. The computational complexity of JCM increases second-degree polynomially with the number of objects. The experiments demonstrate the robustness of the method.

Key words: Multiple Objects Tracking, Particle Filter, Joint Contour Method, Active Shape Model