Calibration and 3D modeling from pan-tilt-zoom camera networks

Principal Investigator: Marc Pollefeys 
Funding Agency: Siemen's Medical Systems, Inc. 
Agency Number: N/A

Abstract

One graduate student at UNC-CH will work on the following under the supervision of Prof. Pollefeys. The goal of this project is to explore flexible approaches to perform calibration and 3D reconstruction of events observed by multiple network pan-tilt-zoom cameras. For calibration we will take advantage of the active pan-tilt-zoom functionality to recover intrinsic and extrinsic parameters of all the cameras in the network (including variation of radial distortion and other parameters in function of the zoom settings). Advanced calibration procedures that take advantage of dynamic components of the scene for calibration and possibly synchronization will also be explored. For 3D modeling we also intend to take full advantage of the flexibility offered by an active network of pan-tiltzoom cameras to model dynamic events taking place in large environments at high resolution (e.g. airport terminal, stadium, public lobby, market square, ...). Our goal is to achieve similar or better results than approaches that use fixed camera networks that are restricted to small working volumes. Note: the student will have access to a network of 6 or more pan-tilt-zoom cameras as well as to all necessary computer infrastructure.