Semantic Traffic Light Understanding for Visually Impaired Pedestrian

摘要

A visually impaired pedestrian normally requires information of walking/stop signs with time to decide whether it is available to cross a road on crosswalk. Since traffic lights are installed in verity formats such as, man figure, color and numerical characters, it is complicate to simply understand the meaning. This paper proposes a method of semantic traffic light understand for a visually impaired pedestrian. In the method, a picture of traffic light is initially assumed to extract and type of traffic light are first classified into man figure, color and numerical characters by some classification tools such as signature. In man figure case, color and man gesture are classified and both information are decided base on Fuzzy logic. In the case of numerical characters, they are first enhanced and then classified by a classifier. The experiments have been performed with 200 samples of pedestrian traffic lights and results reveal acceptable accuracy.