This paper proposes an obstacle detection method for indoor mobile robots using axial stereo camera model. Since the axial stereo model has large viewfield and small occlusion, its correspondence ambiguity is significantly smaller than that of other stereo camera model. In navigation of a mobile robot, it is important to detect the obstacles in a very short time.
In order to reduce the processing time, this paper utilizes the bottom edges of the obstacles which are contacted with the floor. To acquire the range information of the bottom edges, an image is projected on the floor by using the inverse perspective mapping. Then, this projected image is mapped perspectively to a known image plane. This principle enables to find the bottom edges of the obstacles.
Using the correlation method, whether the bottom edges are borders between the obstacles and the floor or not is verified. The verified bottom edges are transformed to the mobile robot's coordinate system in order to make the 2-dimensional map. The effectiveness of this method is shown by a series of experimental results.