基于形状标记图和Gabor 小波的交通标志识别.pdfVIP

11 1 基于形状标记图和Gabor 小波的交通标志识别 1 1 1 谷明琴 ，蔡自兴 ，何芬芬 （1. 中南大学信息科学与工程学院，湖南长沙，410083 ） 摘要：交通标志识别为智能车辆行驶提供了有价值的道路环境信息。交通标志识别过程如下：1)变换图像 的RGB 像素值来增强交通标志主特征颜色（红，蓝，黄）区域并分割图像，用形态学操作排除噪声点的影 响；2)为了获得完整的交通标志轮廓形状，用灰度图像上的边缘检测和连接算法重构感兴趣区域；3)提取 感兴趣区域的特征图作为其形状特征，用Euclidean 距离来对其特征进行分类；4)对交通标志感兴趣区域的 灰度图像进行Gabor 小波变换，获得其不同角度和尺度的小波图像，用二维独立分量分析法提取其主特征， 并送入线性支持向量机来判断感兴趣区域所属的交通标志类型。实验结果表明本文提出的算法能够稳定， 有效的检测和识别智能车辆行驶环境中的多类交通标志。 关键词：交通标志识别；标记图；Gabor 小波；支持向量机 中图法分类号：（作者本人填写） 中图分类号：（作者本人填写） 文献标识码：A Traffic Sign Recognition Based on Shape Signature and Gabor Wavelet 1 1 GU Ming-Qin , CAI Zi-Xing (1.Schoolof Information Science and Engineering, Central South University, Changsha 410083, China) Abstract: Traffic sign recognition provides valuable information of road conditions for intelligent vehicles. Traffic sign recognition process as following: 1) The main colors of traffic signs were enhanced by transforming the RGB pixel values of image and it was segmented by a threshold. Image morphological processing filtered noise points of binary image. 2) In order to obtain complete outline of the shape of traffic signs, Regions of interest (RoIs) were reconstructed by edge detection and connection algorithm on the gray image. 3) Signature of RoI was extracted as its shape feature. This feature was classified by Euclidean distance. 4) Gray image of traffic signs RoI was transformed to different orientations and scales wavelet image by Gabor wavelet, and its main feature was