Probabilistic Voxel Mapping Using Stereo Matching Confidence

Abstract

There has been meaningful research into the development of 3D world modeling techniques that are important requisite for intelligent vehicle navigation. In this paper we describe a 3D probabilistic voxel mapping process using stereo matching confidence. Proposed 3D probabilistic voxel map is more reliable representation than general voxel map that just contains the occupancy information. To get the matching confidence value, we evaluate stereo matching costs and its propagation. We test the proposed method in large-scale outdoor environment.