Robot navigation based on pseudo-binocular stereo vision and linear fitting

Abstract

In this paper, we propose a pseudo-binocular stereo vision and linear fitting to achieve robot navigation. Pseudo-binocular stereo vision is realized by moving the robot body to the left or right with a certain distance. Then, a disparity map is calculated with the semi-global matching algorithm. This method can overcome the interference of complex color and texture for objects that are at the same distance. After edge detection and binarizing of the disparity map, road lines are extracted by the Hough transform algorithm. To extract the road lines from the noisy edges, the least square method for linear fitting is utilized for high-accuracy navigation.