Mobile robot 3D map building based on laser ranging and stereovision

Abstract

In this paper, an effective 3D map building approach based on range data from binocular stereo vision sensor and Laser Range Finder is introduced in detail. First of all, a local map temporal integration approach in which Bayesian filter based dynamic occupancy grid map modeling technique is employed to reasonably deal with measurement uncertainty involved in environment perception. In addition, as stereo vision is unreliable for building map, an effective combination approach is proposed for fusing local stereo data derived map and laser range data derive map together. Finally, a reliable 3D spatial model is built using textures extracted from images. The effectiveness of our proposal is validated by real experimental results carried on Pioneer robot.