摘要： 针对单纯利用RGB或HIS空间不能很好地检测出交通标志这一问题，提出一种多通道融合的交通标志检测方法。将图像转换至HIS空间，使用迭代的Otsu算法对饱和度进行阈值分割，以色相和饱和度信息为基础，结合RGB通道的颜色特征进行颜色聚类。利用水平和垂直投影获得交通标志的外接矩形，利用曲线拟合的最小二乘法定位交通标志并对其进行分类。研究结果表明，该方法可以准确地从背景图像中提取出交通标志，有效地解决存在亮度变化、噪声干扰、方位倾斜等情况的交通标志检测问题。

关键词: 交通标志检测, 颜色通道, 阈值分割, 曲线拟合

Abstract: Aiming at the disadvantages of detecting traffic sign with RGB or HIS color space values respectively, a novel traffic sign detection method with multi-channel integration is proposed. The HIS color space values are converted from RGB values through image color space transformation. Sub-variable S in image segmentation is processed based on the iterative Otsu threshold value algorithm. The color is clustered based on the information of hue and saturation, combining with the color-character of RGB channel. The circum-rectangle is got with horizontal and vertical projection. It allocates and assorts the traffic sign with the least square method of curve fitting. Experiment shows that the method can abstract traffic sign from background precisely, even under the circumstance of brightness changing, noise interferences or azimuth obliquity.

Key words: traffic sign detection, color channel, threshold segmentation, curve fitting

中图分类号: 

• TP309