KNOWLEDGE EXTRACTION FROM SURVEILLANCE SENSORS
RAMA CHELLAPPA, ASHOK VEERARAGHAVAN, AND ASWIN C. SANKARANARAYANAN
Center for Automation Research and Department of Electrical and Computer Engineering, University of Maryland, College Park, Maryland
In the last decade, surveillance and monitoring has become critical for homeland security. With this increasing focus on surveillance of large public areas, a traditional human-centric surveillance system where a human operator watches a bank of cameras is largely being supported with automated surveillance suites. In a typical public area such as an airport or a train station there could be anywhere between 50 to a few hundred cameras deployed all over the area. It is almost impossible for human operators to keep a close watch on all of these cameras and continuously and robustly identify subjects and events of interest. Therefore, there is a greater need for automated analysis of the data obtained from multiple video cameras and other sensors that are distributed all around the area of interest.
Several current state-of-the-art surveillance systems work in aid of human operators. These surveillance systems have a host of sensors [visual, audio, infrared (IR) etc.] that are distributed in the area of interest. These sensors are in turn networked and connected to a central command center, where sophisticated algorithms for varied tasks such as person detection, tracking, and recognition; vehicle detection and classification; ...