中文摘要：

干扰区域排除是交通标志检测的难点。本文对颜色规格化后的图像进行二值化，讨论了二值化过程中感兴趣颜色的选取，得出了以红、蓝、黄为感兴趣颜色引入干扰较少的结论。根据我国交通标志的形状特征提出了两个基于重心的干扰区域判定规则和一种基于道格拉斯一普克算法的交通标志最外层轮廓多边形拟合方法，该方法通过判断拟合多边形的凸性来区分干扰区域与交通标志候选区域。本文算法可以检测我国任意形状的交通标志并对交通标志倾斜及旋转的情况具有较好的鲁棒性。

英文摘要：

Interference regions exclusion was a difficulty to traffic sign detection. This paper discussed the selection of interested colors in the binarization process of color standardized images. After discussion, the conclusion was that taking red, blue and yellow as the interested colors would introduce less interference. According to the shape characteristics of the traffic signs, this paper proposed two decision rules to interference regions based on barycentre and a polygon fitting method to the outer contour of traffic sign based on the Douglas-Peucker algorithm, which distinguished between the interference region and the traffic sign candidate region by the convexity of the fitted polygon. The proposed algorithm could detect any traffic signs of our country and had good robustness to tilted and rotated traffic signs.