Visual input

A CCD camera gives pictures of the environment. High curvature points (corners ...) are extracted from the gradient of the image. In a navigation context, the correspondence between a salient point (called now ``landmark'') with its angular position (azimuth) in respect with an absolute direction (north given by a compass for instance) gives the position of that landmark in the environment. At each time step, a set of 5 azimuths, corresponding to the 5 more salient landmarks in input the image, constitutes the visual input $\mathbf{I}^{(1)}$ (the angular confidence interval is $4^\circ$, so that the number of visual entries is $N^{(1)}=90$).