Topological map from only visual orientation information using omnidirectional cameras

In this paper we present a new way to compute a topological map using only orientation information. We exploit the natural presence of lines in man-made environments in dominant directions. We extract all the image lines present in the scene acquired by an omnidirectional system composed of 6 aligned cameras. From the parallel lines we robustly compute the three dominant directions using vanishing points. With this information we are able to align the camera with respect to the scene and to identify the turns in the trajectory. Assuming a Manhattan world where the changes of heading in the navigation are related by multiples 90 degrees. We also use geometrical image-pair constraints as a tool to identify the visual traversable nodes that compose our topological map. Experiments with an indoor sequence have been performed to validate this approach.