We propose a robust tracking and monitoring system based on a high frame rate, stereo vision 3D sensor combined with color video data. The envisioned system is capable of detecting people or vehicles moving in a large facility - indoors or outdoors, and keeping track of each target individually in real time as they pass from one monitored view to another despite crowds, partial occlusions or lighting changes. A prototype tracking system will be demonstrated, extended and evaluated for performance using existing high-performance stereo correspondence hardware. The envisioned system will be structured to serve as a platform for acquiring 3D data to be integrated with other biometric methods for identification and threat classification. Finally, issues regarding practical field use such as cost, installation and maintenance will be identified and addressed. The Department of Defense (DoD) is at a heightened state of awareness since the terrorist attack on September 11, 2001. Military and civilian facilities need technology that will help them monitor and track potential terrorists before an attack takes place. Existing identification technology and security systems, based on single image sensors or basic motion sensors, are not sufficient to fulfill this need because they are not robust with respect to real world environments. 3D technology though recognized as a invaluable tool for surveillance tracking and identification has been too slow, expensive and cumbersome to incorporate into commercially practical security systems. The results of this SBIR topic will enable practical security systems using this critical 3D modality. When completed, the result will be an entirely new paradigm for visual surveillance cameras and capabilities - away from simple video capture and towards security automation, real time threat analysis and intelligence gathering.
Keywords: Stereo Vision, 3d Tracking, Person Tracking, Head And Hand Tracking, Cmos Image Sensors, Video, Surveillance, Object Tracking
