Computer Engineering and Applications ›› 2008, Vol. 44 ›› Issue (5): 55-57.

• 学术探讨 • Previous Articles     Next Articles

Scene model based object tracking method across multiple cameras

DENG Ying-na,ZHU Hong,LI Gang,LI Lu,QIAN Hui-fang

  

  1. Digital Image Process Lab,Xi’an Unversity of Technolog,Xi’an 710048,China
  • Received:2007-06-07 Revised:2007-08-22 Online:2008-02-11 Published:2008-02-11
  • Contact: DENG Ying-na

基于场景模型的多摄像机目标跟踪算法

邓颖娜,朱 虹,李 刚,黎 璐,钱慧芳   

  1. 西安理工大学 自动化与信息工程学院 图像处理研究室,西安 710048
  • 通讯作者: 邓颖娜

Abstract: In the surveillance system with multiple cameras,camera handoff is the key problem when an object left a view and entered another.This paper addressed this problem by comparing the real world positions of objects.Position in real world is obtained through an environment model which established the correlation of object position in real world and image plans.The edges of the overlapped view are obtained through a man walking in camera views freely.Then the system tracked object continuously across multiple cameras.This method accommodates multiple objects entering the view simultaneously across multiple arbitrary equipped cameras and the experimental result show is robustness and efficiency.

Key words: multiple cameras, object tracking, position acquiring, overlapped view

摘要: 在多个相机组成的视频监视系统中,当目标物移出某一相机的视野而进入下一个时,如何实现相机的交接,实现目标物的继续跟踪是监视系统中要解决的关键问题。针对该问题,提出了一种基于位置比较的多摄像机运动目标跟踪方法。为获得目标物的位置,建立多个相机与目标物世界坐标之间映射关系的场景模型,并根据目标物出现在不同相机之间的视野边界线上的瞬间时刻的位置来给出重叠视野的边界线。由此可对任意角度摆放的多个具有重叠视野的相机之间运行的目标物进行接力跟踪。该方法可以适应多个目标物同时进入场景的情况,实验结果表明,该方法具有较高的鲁棒性,能够满足视频跟踪的实时性要求。

关键词: 多摄像机, 运动目标跟踪, 位置信息的确定, 重叠视野