Object Tracking With A PTZ Controlled Camera

Nowadays, the desire for safety in public locations, such as airports and governmentbuildings where many people gather, has been increasing. As a result, these places usuallyadopt video-based surveillance, so intelligent visual surveillance has become anextensively researched subject in computer vision community. Object detection andtracking is a key technique in computer vision and security surveillance.The traditional object tracking methods can’t track the object in a large area and along time, which is very important for intelligent analysis, for their field-of-view (FOV) islimited. To enlarging the monitoring area, more and more visual surveillance systemsexploit Pan Tilt Zoom (PTZ) camera.As the movements of camera, the background changes dynamically, much newbackground is added in. Background difference methods have been improved, by manyresearchers, to solve this problem, such as mosaic background method and backgroundcompensation method. However, these methods are high computational complexity for theadditional processing of background movements.Firstly, the advantages and limitations of object detection and tracking have beenanalyzed. Then, an object tracking method with a PTZ controlled camera is proposed,which is based on candidate area matching. Objects are represented using color andtexture information, and their locations in images, which are used to automatically controlthe camera’s PTZ, are found out by template matching. Since the delay of transferring andexecuting the commands of camera, the location of the object after the delay time ispredicted through the trajectory of the object. Experimental results show that the objectcan be tracked in large scenes by our method, even though the object are occluded after ashort time, and the object is almost always localized within the field-of-view of thecamera.