Panoramic background generation and abnormal behavior detection in PTZ camera networks

Abstract

In this paper, we present a novel method to abnormal behavior detection in PTZ camera networks. To extract motion information of scene in moving camera environment, we use panoramic background which is generated by frames. In contrast with previous methods, we use MRF framework to integrate temporal and spatial information of frames to generate panoramic background. In addition, we introduce a panoramic activity map to detect abnormal behavior of people. The panoramic activity map is useful in various abnormal situation since it includes multiple features such as location, motion, direction and pace of objects. Experimental results from real sequences demonstrate the effectiveness of our method. © 2011 Springer-Verlag.