Orb-Slam Examination in the Indoor Textureless Environment

Abstract

Visual SLAM is a challenging topic because the algorithm needs to be run in real-time and give a precise estimation of camera pose. The field of research become popular after Davison [1] introduced computer vision methods to SLAM. The current state-of-the-art SLAM algorithms: LSD-SLAM [2] and ORB-SLAM [3] can provide camera trajectories and build a precise map. However, these research have few data and experiments for the indoors. The research would focus on applying SLAM methods in the indoor environment because of the rising demands of indoor robots. The purpose of the research is to evaluate the performance when ORB-SLAM algorithm works in a textureless indoor environment. ORB-SLAM is believed to be one of the best among state-of-the-art visual SLAM algorithms, where it adopts a heuristic model selection to tackle with different scene scenario in their experiments. However, the researcher is interested in whether repetitive patterns, textureless walls, and planar structures in the indoors affect the robustness of a visual SLAM algorithm. The research conduct experiments in the HRBB 4th floor to examine the performance of ORB-SLAM in a textureless indoor environment.