Visual tracking based on multiple cameras with MHS cooperation strategy

In this paper, multiple cameras are used to implement a multi-purpose real-time visual tracking system. A modified adaptive background subtraction method is used to detect moving objects, and a multi-cue matching approach is employed to track a moving target. The epipolar constraint and color histogram matching are used to deal with the correspondence problem that occurs when multiple cameras are used without camera calibration being performed. An effective and robust Master-Helper-Slave (MHS) cooperation strategy is proposed to further improve the tracking performance. An active PTZ camera is integrated to perform dynamic tracking. The proposed visual tracking system can continuously track a moving target and estimate its corresponding position in the field of view (FOV) of each camera, even if the target is occluded in some of the cameras' FOVs. Experimental results show that the proposed approach exhibits satisfactory performance.