Depth map generation for a reconnaissance robot via sensor fusion

Abstract

In this paper design and implementation of an accurate positioning system in indoor environments for an Omni directional Robot and a stereo vision based depth estimation system is discussed. The positioning system which is robust for wheel slip and jerky motion was implemented fusing odometry, Inertial Navigation System (INS) and magnetic compass based measurements. The vision system was implemented using two off the shelf web cameras. Here, Semi-Global Block Matching (SGBM) method was further developed considering multiple realizations of the same scenery to increase the accuracy of the depth estimation.