Real-time Arrow Traffic Light Recognition System for Intelligent Vehicle

Arrow traffic lights are common at intersections of current urban scenes. However, existing algorithms mainly focus on recognition methods of circle traffic light. A novel system used to detect and recognize arrow traffic lights in urban scenes is proposed in the paper. Firstly, the boards of traffic light are exactly localized by image segmentation and morphology processing, then candidates of traffic light are obtained according to threshold segmentation in YCbCr color space and judged by relative position between candidate and its board. Secondly, for confirming traffic light type, Gabor wavelet transform and 2D independent component analysis(2DICA) are used to extract traffic light candidate’s features. A nearest neighbor classifier identifies arrow direction and removes noise. Experimental results show that overall recognition rate of proposed method exceeds 91%, and system will provide real-time, robust, effective, and accurate traffic light outputs for intelligent vehicle.