Real-time walk light detection with a mobile phone

Abstract

Crossing an urban traffic intersection is one of the most dangerous activities of a blind or visually impaired person's travel. Building on past work by the authors on the issue of proper alignment with the crosswalk, this paper addresses the complementary issue of knowing when it is time to cross. We describe a prototype portable system that alerts the user in real time once the Walk light is illuminated. The system runs as a software application on an off-the-shelf Nokia N95 mobile phone, using computer vision algorithms to analyze video acquired by the built-in camera to determine in real time if a Walk light is currently visible. Once a Walk light is detected, an audio tone is sounded to alert the user. Experiments with a blind volunteer subject at urban traffic intersections demonstrate proof of concept of the system, which successfully alerted the subject when the Walk light appeared. © 2010 Springer-Verlag Berlin Heidelberg.