Fractal representation of image feature associated with maneuvering affordance

One of the essential capabilities of 'real world intelligence', whether developed naturally or designed artificially, is to generate feasible operations based on innate belief in real world. As cognitive basis of the real world intelligence, visual perception organizes randomly distributed image features into environment features: well structured visibles available as consistent cues to subsequent decisions. Such phenomenal supervenience to reality plays a crucial role in implementing cooperative systems intended for field automation, vehicle-roadway networking, community restoration from disaster, and interactive education, e.g. in generating consistent decisions, partial knowledge of the environment should be adapted intentionally to encountered scene prior to the comprehension of the situations. Such selfreference structure, however, yields serious contradiction in understanding natural perception mechanisms and/or implementing artificial vision systems. In this paper directional Fourier transform was applied to extract maneuvering affordance in noisy imagery. By identifying the brightness distribution of observed patterns with the invariant measure of unknown fractal attractor, noise levels were estimated for extracting affordance pattern. The detectability of affordance patterns has been verified through experimental studies.