Traffic Sign Detection Based on Fixed Color Combination and Intensity Restraint

The traffic sign detection problem is investigated in this paper. A segmentation method for traffic sign image in HSV color space is presented for the detection of prohibition sign, warning sign and mandatory sign. First, the image is pre-segmented by H and S components with threshold value in HSV space. Second, the regions that their area and the ratio of width and height are outside the threshold value will be removed. Finally, the traffic sign final segmentation is performed and its location is detected, in which only the regions with the color and background being matched with the given color combination and the intensity requirements are preserved. The experimental results show that three kinds of traffic signs can be detected quickly and accurately for real datasets.