On augmenting the visual slam with direct orientation measurement using the 5-point algorithm

Abstract

This paper presents the attempt to merge two paradigms of the visual robot navigation: Visual Simultaneous Localization and Mapping (VSLAM) and Visual Odometry (VO). The VSLAM was augmented with the direct, visual measurement of the robot orientation change using the 5-point algorithm. The extended movement model of the robot was proposed and additional measurements were introduced to the SLAM system. The efficiency of the 5-point and 8-point algorithms was compared. The augmented system was compared with the state of the art VSLAM solution and the proposed modification allowed to reduce the tracking error by over 30%.