Surface Component Ratio Histogram for RGB-D SLAM in Indoor Environments with Low-Textured Scenes

Abstract

RGB-D sensor based 3D perception and simultaneous localization and mapping have been widely investigated in recent years, but still there are remaining challenges to ensure its performance in low-textured environments. The depth image is useful in dealing with such issues. In this study, we propose depth based surface-level feature extraction for the SLAM scheme that does not rely on image textures. Unless the scene is provided with sufficient image textures, our approach can be an alternative solution to the feature association during SLAM.