Abstract: How to detect traffic sign from real scene is a key step in traffic recognition system.According to the color and shape characteristics of Chinese speed limit traffic signs，this paper proposes a new kind of filtering approach that is used to obtain an elliptical model where red pixels cluster in Lab color space.Then the elliptical model is used to extract area of red pixels and gain the binary image of traffic sign.After that，the improved Hough transform is used to detect elliptical region of binary image of traffic sign.The result comes true an intelligent speed limit sign detection method that combines the priori characteristics of the traffic sign and machine learning algorithm.

Key words: speed limit sign, priori characteristics, AdaBoost, Hough transform

摘要： 如何从实景中有效地提取出交通标志是交通标志识别系统的关键，在分析中国道路限速交通标志的颜色和几何形状两种先验特征的基础上，以一种新的颜色滤波方法为基础，获得红色像素在Lab颜色空间中聚类范围的椭圆模型，提取出图像中的红色区域，得到二值化图像，然后采用基于梯度信息的Hough变换圆检测方法获得二值化图像中的圆和椭圆区域，从而实现一种将交通标志先验特征与机器学习算法相融合的智能检测方法。

关键词: 限速标志, 先验特征, AdaBoost算法, Hough变换