A Fast Algorithm for Finding Crosswalks using Figure-Ground Segmentation

Urban intersections are the most dangerous parts of a blind person’s travel. Proper alignment is necessary to enter the crosswalk in the right direction and avoid the danger of straying outside the crosswalk. Computer vision is a natural tool for providing such alignment information. However, little work has been done on algorithms for finding crosswalks for the blind, and most of it focuses on fairly clean, simple images in which the Hough transform suffices for extracting the borders of the crosswalk stripes. In real-world conditions such as cluttered scenes with shadows, saturation effects, slightly curved stripes and occlusions, the Hough transform is often unreliable as a pre-processing step. We demonstrate a novel, alternative approach for finding zebra (i.e. multiplely striped) crosswalks that is fast (just a few seconds per image) and robust. The approach is based on figure-ground segmentation, which we cast in a graphical model framework for grouping geometric features into a coherent structure. We show promising experimental results on an image database photographed by an unguided blind pedestrian, demonstrating the feasibility of the approach.