Multi Camera Multi Object Tracking using Block Search over Epipolar Geometry

• Main Authors: Saman Sargolzaei, Nilesh Patel
• Format: Article
• Language: English
• Published: Institute of Technology, Nirma University 2000-01-01
• Series: Nirma University Journal of Engineering and Technology
• Online Access: http://nujet.org.in/index.php/nujet/article/view/97

We present strategy for multi-objects tracking in multi camera environment for the surveillance and security application where tracking multitude subjects are of utmost importance in a crowded scene. Our technique assumes partially overlapped multi-camera setup where cameras share common view from different angle to assess positions and activities of subjects under suspicion. To establish spatial correspondence between camera views we employ an epipolar geometry technique. We propose an overlapped block search method to find the interested pattern (target) in new frames. Color pattern update scheme has been considered to further optimize the efficiency of the object tracking when object pattern changes due to object motion in the field of views of the cameras. Evaluation of our approach is presented with the results on PETS2007 dataset..