作者: Sudipta Sinha , Marc Pollefeys
DOI:
关键词:
摘要: In this paper we discuss the problem of recovering calibration a network pan-tilt-zoom cameras. The intrinsic parameters each camera over its full range zoom settings are estimated through two step procedure. We first determine at camera’s lowest setting very accurately by capturing an extended panorama. Our model includes radial distortion. intrinsics and distortion then determined discrete steps in monotonically increasing sequence that spans Both fully automatic do not assume any knowledge scene structure. validate our approach calibrating different types pan tilt cameras placed outdoor environment. also show high-resolution panoramic mosaics built from images captured during process. present for accurate computation epipolar geometry based on panorama instead individual image pairs. Finally, briefly how can be used to compute extrinsics all context active networks.