Real-time arrow traffic light recognition in urban scenes

A novel approach for detecting and recognizing arrow traffic lights in urban scenes was proposed. Firstly, the boards of arrow traffic lights were localized by image segmentation and morphology processing. Secondly, color image of traffic light board region was converted to YCbCr color space. Candidates of traffic lights (red, yellow, green) were obtained through threshold segmentation in Cb and Cr channels, by judging morphology and relative position between candidate and its board. Thirdly, Gabor wavelet transform and 2D independent component analysis (2DICA) were used to extract traffic light candidate's features, and finally nearest neighbor classifier identifies arrow direction. Experimental results indicate that the overall recognition rates of the proposed method are over 91%, and computation time of each frame is 152 ms. So the proposed algorithm will provide real-time, robust and accurate arrow traffic lights information to moving vehicles.