A Novel SLAM Algorithm based on Region Focusing Image Segmentation

Abstract

In order to solve the problem that tracking drift and localization failure caused by pose estimation error in orb-slam (simultaneous localization and mapping), an improved orb feature point extraction algorithm called RFIS-slam(Region Focusing Image Segmentation slam) is proposed. Firstly, when extracting feature points, the image is divided into several areas and feature points which to be extracted is equally distributed into each area to improve the feature points to be more representative. Secondly, an evaluation system is added to evaluate each area, and some areas with low contrast that are not easy to extract highly identifiable feature points are excluded to obtain more robust feature points. Finally, the experimental results were verified that the RFIS-slam improves the feature point matching by 4.1% and 5.8% respectively compared to orb-slam and orb-slam2; in terms of initialization, 111 and 65.4 frames were improved respectively. Better results have also been achieved in position estimation and real-time positioning, so as to improve the precision of feature point matching and positioning without losing speed.