Bio-inspired visual self-localization in real world scenarios using Slow Feature Analysis - Fig 1

Simulated rotation for (a) simulator and (b) real world experiments. The circular image of the surrounding is transformed to a panoramic view with periodic boundaries. Rotation is simulated for every view from one location by laterally sliding a window over the panoramic image with increments of 5 pixels. Thus the variable φ denotes the relative orientation w.r.t. the robot’s global orientation. Arrows indicate a relative orientation of 0°, 90°, 180° and 270°. (c) Due to the periodic image boundary we can simulate a full rotation with the sliding window approach. The part of the image covered by the window represents the data that is processed at one time step. The size of the sliding window is given as the percentage of the original panoramic view.