Two-step calibration algorithm to estimate the extrinsic parameters of the omnidirectional camera

초록

The omnidirectional camera system with its wide view angle is widely utilized for camera calibration and 3-D reconstruction due to its advantage that a large amount of environmental information can be obtained using a small number of images. Since the straight-line features of the environment are projected as contours by the omnidirectional camera model, the corresponding contours among views can be effectively used for camera calibration. This paper presents a two-step calibration algorithm to estimate the extrinsic parameters of the omnidirectional camera by using the corresponding contours. The first step involves obtaining camera parameters by minimizing an angular error function between the epipolar plane and back-projected vectors of end points on the corresponding contours. With this, the final parameters minimizing a distance error between the computed and the real contours in the third view are estimated. Our experiments on synthetic and real images demonstrate that the algorithm precisely estimates the extrinsic parameters of the camera. (C) 2008 Society of Photo-Optical Instrumentation Engineers.

키워드