Algorithm On Panoramic Collaborative Target Tracking With Multiple Cameras

With social development, intelligent video surveillance technology experienced manpower on-site monitoring, video surveillance and post-human intelligent video surveillance has been rapid development. Initially, the eyeshot of single camera is limited, can’t effectively solve the problems of the continuous tracking and targets occlusion. Later, using more than two collaborative surveillance cameras, increasing the viewing angle, you can expand the camera’s field of vision field, capturing moving targets comprehensive information. Helps to solve the target block, environment illumination mutations situation, the3d object modeling and so on. Interaction between multiple cameras would be develop two genres: non-overlapping region tracking and overlapping region tracking.Combined with respective characteristics of overlapping and non-overlapping field-of-views effectively, this paper proposes a multi-view layout construction method and a panoramic collaborative tracking algorithm with multiple cameras. Our work contents are as follows:First, using double vanishing point perspective correction principle commonly calibrates multiple cameras to unify the global coordinate system, and fusing multi-view images information establishes a panoramic view. Second, using mixture gaussian models to segment foreground objects, and using mixture gaussian models to eliminate shadows. Further post-processing for the prospects gets the target destination domain rules. Third, combining with the saliency of overlapping target domains in panoramic view and the single view tracking information achieves to collaborative targeting of the target. At last, the panoramic view also provides the trajectory for global monitoring.Compared with other algorithms, the advantages of the proposed algorithm include: solve the small field of view of the lack of a single camera, combine with non-overlapping and overlapping the respective advantages of sight, to achieve global tracking, simple calibration procedure, availability to large scenes and small target tracking, and more useful multi-camera layout.