Stereo vision SLAM system based on ORB

Abstract

Vision-based Simulation Localization and Mapping (VSLAM) is divided into front-end and back-end, front-end including visual odometer and loopback detection, and the back-end including optimization and map-building using stereo camera, this paper by using calculation of parallax to calculate image pixel depth, by using the ORB corners to realize feature points extraction and matching, and then apply graph optimization to implement local map optimization, by using Bag-of-Words (BoW) module to complete loopback detection, finally, this paper realizes the ORB angular point-based sparse landmark map. This paper provides the realization of the stereo vision based on ORB, and then discuss the direction of further development on SLAM.