Real-time video surveillance for large scenes

Abstract

Video surveillance provides security monitoring of target scenes. As the public safety is confronted of more and more neoteric challenges, numerous video surveillance cameras are deployed around the world for public safety and security, intelligent traffic management and homeland security. With the advances in both hardware and software, video surveillance systems have been innovated with many useful features, such as facial recognition, and integrated with networking technologies. However, constrained by optical technologies and costs, most cameras used in practice are designed to provide only limited angles of view, namely partially cover a scene, especially when it is large. Moreover, most of the cameras installed in current video surveillance are analog closed circuit television (CCTV) cameras that do not support IP capability for remote control, easy scaling in deployment and further mining of video information. This paper introduces a scalable multi-tier economic video surveillance solution with a unique feature-realtime video stitching. This system takes advantage of advance in networking technologies, cloud computing and computer vision. Preliminary studies have shown the merits of this solution.