DETECTION OF DOMINANT PLANAR SURFACES IN DISPARITY IMAGES BASED ON RANDOM SAMPLING

Abstract

In this paper, the applicability of RANSAC-approach to planar surface detection in disparity images obtained by stereo vision is investigated. This study is specially focused on application in indoor environments, where many of the dominant surfaces are uniformly colored, which poses additional difficulties to stereo vision. Several simple modifications to the basic RANSAC-algorithm are examined and improvements achieved by these modifications are evaluated. Two simple performance measures for evaluating the accuracy of planar surface detection are proposed. An experimental study is performed using images acquired by a stereo vision system mounted on a mobile robot moving in an indoor environment.