Traffic Sign Detection Method Using Multi-Color Space Fusion

Many traffic sign detection algorithms suffer from poor robustness and low detection efficiency problems under poor environmental conditions, such as low illumination and sudden weather changes. Here, we propose a novel traffic sign detection algorithm based on multi-color space fusion. First, we select sensitive channels that can represent the color information of traffic signs by analyzing the color space. Second, the sensitive channels from the equalized image are extracted and fused to enhance the contrast of the area with the traffic sign colors. Additionally, an adaptive threshold setting method is used to binarize the fused image to acquire the candidate areas of traffic signs. Finally, non-traffic sign areas in the candidate areas are removed on the basis of the features of area, aspect ratio, and eccentricity. Results using two databases demonstrate that our proposed algorithm can effectively improve the detection rate and has a higher robustness under poor environmental conditions.