In this study, we design a dual cooperative camera system to maintain the both requirements of a visual surveillance system of tracking and close-up viewing functions in real time. In most of the object tracking or detecting systems, the details or close-up views of the objects are easily neglected for surveilling the environment to keep the camera in wide field view. Besides, some of the systems use the master-slave mode to keep one camera always in a close-up mode and the other camera in searching mode, and it may cause the dead end problem in surveillance or miss the target for not allowing the dynamic change of the roles of both cameras. Therefore, the cooperative dual camera system can not only extract the object correctly by analyzing the human body's shape but also predict the moving position and height of the human to have a close-up view. The dual pan-tilt-zoom (PTZ) cameras are used in the system to have the adjustable zooming and focusing parameters. The experiment results demonstrate that the proposed algorithm can obtain clear close-up views for either a single target or multiple objects in real time.