Closeup-Imaging of a Moving Object by Cooperative Multiple Pan-Tilt Cameras
Hicham Bouchnaif · 2005
In this thesis we present a framework for intelligent filming of an object moving in a wide field. We propose a new method of filming of the moving object by fixed framing of multiple pan-tilt cameras. This filming approach assures the acquisition of high resolution images of the object without blurring of its textures, which may occur when a conventional active tracking method is applied. Since the camera that is filming the object cannot be rotated in fixed framing approach, cameras which are not used for filming at that time should be directed beforehand so as to maximize the possibility of capturing the object by their fixed framing pose. To realize this system, we decouple the framing task into two parts, a monitoring part, and a stand-by part. We assign a monitoring role to one camera, and a stand-by role to other cameras. While the monitoring camera films the target, the motion information of this target is embedded into a marginal distribution through the projection of a probability density function onto the circumference of the image plane. Based on this new marginal distribution, the stand-by cameras predict the moving direction of the target. Each of these