Object tracking system by adaptive pan-tilt-zoom cameras and arm robot

This paper proposes an object tracking system using an arm robot and pan-tilt-zoom cameras. The object detection method consists of the sparse template matching using color information and the particle filter methods for a fast detection. And estimate 3D position of it by the binocular stereo method using two cameras. In order to detect object stably, we control the zoom (the focal length) of cameras adaptively considering 3D object positions. We configure the object tracking system and confirm the effectiveness of proposed method in experiments that the object moves surround of this system.