Abstract
A stochastic scheme is presented for the identification of a roadway pattern in complex natural scenes. By matching linear scale shift rule with randomly distributed noise images, a version of invariant measure is estimated as the observation of not-yet-identified fractal attractor spanning the roadway area within a generic ground-object structure. The expansion of the invariant measure is restricted by associated breakdown pixels with respect to the linear scale shift rule to design the self-similarity process generating the attractors. The consistency of designed process is verified via invariant feature detection and visualization on the scene images. Proposed method is verified through experimental studies using various types of roadway scene images.
