Traffic light detection based on similar shapes searching for visually impaired person

Abstract

A new method of circular traffic light detection and recognition was proposed for the visually impaired person to help them cross the road. Similar circular and rectangular shapes detection combined with the intensity analysis was used to locate the candidate traffic lights. And traffic light color recognition was realized using color distributions of the light area in HSI color space and shape features of the lamp. Voice prompt was given after analyzing the recognition results of video sequences. Experimental results showed that the traffic light detection and recognition method has high recognition rate of 98%, and the frame processing speed was 0.09s, which met the real-time requirements.