Detection of flat ground area by single camera on mobile vehicles

摘要

For moving robots of outdoor use, it is essential to detect the movable road area. In this research, we propose a new method of making distinction between road points and off-road points that are detected as feature points. Namely, we are to build the projection transformation model of the road seen from the robot in advance, the robot detects feature points by SURF or some other methods, and then making correspondence between the consecutive frames. By inverse projection, the points can be classified into on-road and off-road points. In this paper, we propose the principle of this theory and show some examples.