Segmentation of Corridor Images for Structure Based Robot Navigation

Abstract

Extracting the structure of space around a robot, is an important issue for robot navigation. In order to identify the structure of a corridor, it is necessary to analyze its composition. We use a camera to take some photos from the corridor in front of the robot, we apply the Sobel algorithm to detect edge lines in two directions, Sobel algorithm finds horizontal and vertical edge lines, the vertical and horizontal edges are respectively related to the wall and floor. then we use a threshold to reduce unnecessary edge linesc and refuse the useless edge lines and eliminate them by introduced threshold methods, so it reduces the computation time and prevent the impact of unnecessary data. We also use the intersection points of the edge lines to obtain the boundary when vertical and horizontal lines cannot give us the wall-floor boundary lines. for more robustness we use a segmentation algorithm based on clustering. These three together help to identify structural cues of corridor such as wall, floor, wall-floor borderline. These cues are necessary for robot navigation in indoor environment.

Keywords

• image segmentation
• corridor structural cues
• robot navigation