ORB-SLAM2S: A Fast ORB-SLAM2 System with Sparse Optical Flow Tracking

Abstract

This paper presents ORB-SLAM2S, a fast and complete simultaneous localization and mapping (SLAM) system based on ORB-SLAM2 for monocular, stereo, and RGB-D cameras. The system works, ensuring accuracy simultaneously, in real-time on standard central processing units (CPU) at a faster speed in small and large indoor and outdoor environments. The system includes a lightweight front-end which is a sparse optical flow method for non-keyframes to avoid the extraction of keypoints and descriptors that allows for high-speed real-time performance. For keyframes, a feature-based method is used to ensure the accurate trajectory estimation almost the same as ORB-SLAM2. The evaluation of famous public sequences shows that our method achieves almost the same state-of-the-art accuracy as ORB-SLAM2 and faster speed performance which is 3~5 times that of ORB-SLAM2, being in most cases the faster SLAM solution. As proved by experiments, the system provides a fast and lightweight visual SLAM while ensuring accuracy for low-cost mobile devices.