Design And Implementation Of Moving Target Tracking Algorithm Based On PTZ Cameras

PTZ cameras are widely used in moving target tracking systems because of theiraflexibility on rotation and zoom. However, researches in the past were based onstationary cameras mostly, and because PTZ cameras rise in recent years, researchesbased on them have many flaws. In view of the effective algorithms are based onstationary cameras, and there are so many flaws in researches based on PTZ cameras,this paper presents the moving target tracking algorithm based on PTZ cameras toresolve this contradiction, applys the algorithms based on stationary cameras to PTZcameras, keeps these algorithms’ robustness and makes full use of PTZ cameras’functions of rotation and zoom at the same time.The moving target tracking algorithm of this paper is divied to three step:hierarchical background modeling, current-frame registration and target tracking. Thehierarchical background model applys a discretization approach to divides theconsecutive focal length scale of PTZ cameras into several discrete level, then breaksup the panoramic scene into many partial fixed scenes at each level. Algorithms basedon stationary cameras can be used in every partial fixed scene. So, PTZ cameras’varied scene is replaced by the collection of partial scenes. Current-frame registrationlocates the current frame. Partical scenes which correspond to the current framecaptured at arbitrary scale and location can be found quickly by using image matchingmethod which is speeded up through three steps to detect foreground. After usinghierarchical background model and current-frame registration, foreground can becombined to tracking algorithms to track moving targets, and then a robust trackingalgorithm is obtained. Using hierarchical background model and current-frame registration, this paperenables applying stationary cameras-based algorithms on PTZ cameras, and trackstargets with the help of foreground, implements the moving target tracking based onPTZ cameras. At last, the implementations prove the robustness of this paper’stracking algorithm.