A Framework for Intelligent Video Surveillance

This paper presents a framework for intelligent video surveillance. Robust distortion correction and rendering algorithms permit the system to support a wide range of video stream sources, including 360-degree panoramic video systems, and to facilitate the substitution of video stream sources, leading to increased scalability. Adaptive object detection, classification and tracking algorithms are employed in a progressive training configuration that allows the system to learn from previous mistakes and to improve performance with increased use. Multiple video stream sources can be combined with arbitrarily shaped overlay surfaces designed to overlay spatially resolved and localized detected and tracked objects on global / local maps such as Global Positioning System (GPS)-enabled maps of the monitored scene for vastly improved situational awareness. Another unique feature of the system is the incorporation of bi-directional communication between the overlay surfaces and the rest of the system via the control unit that enables the use of the localized objects displayed on the overlay surfaces to control the acquisition of higher resolution video corresponding to the region containing the objects in the monitored scene.