Robot navigation by visual local symmetry reflection axes

A method for robot navigation is presented. It is based on the detection of the main local reflection symmetry axes in images where straight lines are prevailing such as structured environments where robots are moving. The objective is to maintain the right orientation during navigation and to detect and correct any change in this orientation comparing continuously the symmetry axes between images. The algorithm measures the positions and the lengths of these local symmetry axes and their relative importance by means of an accumulation map of each contributing segment. A definition of symmetry between two segments that allows us to calculate a local symmetry axis between them is also given.