An integrated scheme of real-time multi-camera tracking for video surveillance
LE3 .A278 2003
Master of Science
In many surveillance applications, multiple video sensors are employed in a cooperative sensor network to provide extended coverage of the monitored site and multiple perspective views of interested targets. This thesis presents an integrated system framework and related algorithms to track moving objects in the monitored site covered by such a sensor network. Three different tracking scenarios, namely single-camera tracking, stereo camera handoff and wide-area tracking, are tackled. Object tracking in a single camera is performed using adaptive background subtraction followed by a heuristic region correspondence strategy. Stereo camera handoff is accomplished using geometric constraints between the coordinate system of the two stereo cameras. Wide-area tracking, i.e., tracking objects from one camera to another, non-overlapped camera is performed utilizing an illuminant-invariant colour-based object appearance modeling/matching algorithm. An integrated system framework is designed for the sensor network involving two main components, the Sensor Processing Unit (SPU) and the Central Control Unit (CCU).
The author grants permission to the University Librarian at Acadia University to reproduce, loan or distribute copies of my thesis in microform, paper or electronic formats on a non-profit basis. The author retains the copyright of the thesis.