An experimental study of a robust self-calibration method for a single camera

This paper presents a robust self-calibration method as well as results from experiments and simulations. The algorithm tracks points in an image sequence and exploits successive views in order to estimate the fundamental matrix and finally the intrinsic parameters of the camera. In the considered application, the camera is mounted on an underwater vehicle in a vertical position. Taking into account the constraints and the controllability of the underwater vehicle´s movements, we study the influence and exploitation of these movements for the estimation of the intrinsic parameters. This paper presents the major steps of the robust self calibration algorithm. Starting from a set of matched points that may comprise bad matches, we describe the estimation of the fundamental matrices using a RANSAC algorithm scheme and the computation of the intrinsic parameters. Experiments are explained and a sensible motion for self-calibration is proposed.