Active multi-camera navigation in video surveillance systems

Active cameras are increasingly used to extend the area coverage and to improve image quality in video surveillance. With such a pan-tilt-zoom camera, operators can control its rotation around two axes and can use an optical zoom to get a better shot. Although multiple video streams are monitored simultaneously on a large grid of displays, they lack spatial correlations that the operators can rely on. In stressful situations, such as during the pursuit of a target, it is hard for the operator to maintain situation awareness while controlling cameras and switching between them. In this work, an interface is presented that helps the operator in controlling the cameras while maintaining spatial knowledge of the environment. Its design extends the work of De Haan et. al. on augmented surveillance interfaces, which only uses static video cameras. First, pan-tilt-zoom camera operation is integrated by providing the user with new camera controls and visual feedback methods. Second, the special hardware joystick is replaced by a simple mouse and keyboard control in order to have a seamless integration with the overall system. Finally, the user is assisted in steering multiple active cameras by automatically determining and focusing an optimal camera based on user-indicated target location and movement direction. As active cameras allow live interaction by the operator, it is difficult to experiment and consistently perform evaluations using real-life surveillance systems. Therefore, the prototype interface is implemented upon a simulator which generates virtual scenarios and live content. Evaluation measures are designed to measure situation awareness and performance of the operator during the pursuit of a target. In the evaluations performed, the prototype was found to have a better learning curve, a better ease of use and user performance is increased in comparison to a classical interface. These findings indicate that the presented interface can enhance operator performance in surveillance control rooms that monitor vast and complex areas.Computer GraphicsComputer ScienceElectrical Engineering, Mathematics and Computer Scienc