Abstract: An improved method of color image segmentation in HSI color coordinate system is presented in this paper.This method combined with the theory of invariant moments is also used for the detection of prohibitory signs in Color Images.First，images are preprocessed，including RGB color coordinate system being transformed into HSI color coordinate system or the improved HSI color coordinate system and then morphological filters being employed.After that，all the closed regions in the image should be signed with different figures and the area of the region which is less than a threshold of the area should be wipe off.The region of which the area is bigger than the threshold should be retained.hu’s invariant moments of every region are calculated and 7 eigenvalues of every region can be obtained.These eigenvalues should be compared with the orthonormal circular and triangular eigenvalues which are prepared in advance.A euclid-distance classifier is designed for comparing.Experimental results show that the prohibitory signs can be detected quickly and exactly in this method.

Key words: road sign detection, improved HSI color coordinate system, invariant moments, euclid-distance classifier

摘要： 提出了一种改进的彩色图像分割方法，并将该方法与不变矩理论相结合用于检测彩色图像中的交通禁令标志。首先对采集图像进行预处理，包括对图像进行RGB到HSI或改进HSI颜色空间的转换和图像形态学的运算，然后对图像中不同的封闭子区域进行标记，并去除不满足面积阈值的子区域。分别计算剩下子区域的hu矩组得到每个子区域的７个图像特征值。将相应子区域的特征值与事先准备好的环形和三角形路标特征值用欧式距离分类器进行比较判别。实验结果表明，此方法能准确并较为快速地实现警告标志检测。

关键词: 道路交通标志, 改进HSI颜色空间, 不变矩, 欧式距离分类器