Epipolar line extraction using feature matching

Abstract

Knowledge of the epipolar lines is of utmost importance in stereo vision since they describe the geometrical relationship between the world points and their projections on the imaging sensors. In this paper we propose a two-phase method to solve the epipolar geometry. Firstly, we extract the corners from a stereo pair and their correspondences are determined by imposing a number of matching constraints. The second phase refines the initial matching results by detecting and eliminating the incorrect matching decisions. The algorithm has been developed for a stereo rig which is used in conjunction with a mobile robot and experimental results are reported.