Using symmetry as a feature in panoramic images for mobile robot applications

We propose to use symmetry as a global feature in panoramic images for mobile robot applications in an indoor environment, where enough man-made symmetrical objects can be found. Our robot system uses an omnidirectional vision sensor consisting of a digital color video camera and a hyperbolic mirror as the main sensor Thus, robust image feature extraction is required for good performance in each application referring to robot navigation. The detection of symmetry is an effective natural vision routine resulting in more global features than edges, corners, etc. Each image can be reduced to certain points of symmetric interest, which can be tracked in all azimuthal directions. A qualitative symmetry algorithm and experiments are presented for range estimation and object classification.