基于感兴趣区域和HOG-CTH特征的交通标志检测

Traffic Sign Detection Based on Regions of Interest and HOG-CTH Features

提出了一种基于感兴趣区域和HOG-CTH融合特征的交通标志检测算法。首先在HSV彩色空间进行颜色阈值分割,然后对分割后的二值图像进行一系列形态学处理获得感兴趣区域,最后提取感兴趣区域的HOG-CTH融合特征,并采用支持向量机(Support Vector Machine,SVM)分类器进行交通标志训练与检测。在特征提取阶段首先分别提取图像的梯度方向直方图(Histogram of Oriented Gradient,HOG)特征和统计变换直方图(Census Transform Histogram,CENTRIST/CTH)特征,然后将CTH特征向量细量化,最后组合HOG特征和稀疏化的CTH特征。实验结果表明,该方法具有很好的鲁棒性,能够快速准确地检测出交通标志。

A traffic sign detection algorithm based on the regions of interest and Histogram of Oriented Gradient and CensusTransform Histogram is proposed. Firstly,the color threshold segmentation is performed in the HSV color space,then a series ofmorphological processing is performed on the segmented binary image to obtain the regions of interest,finally,the HOG-CTH fea-tures of the regions of interest are extracted,and the support vector machine（SVM）classifier is used to carry out traffic sign train-ing and detection.Firstly,the features of the Histogram of Oriented Gradient（HOG）and the Census Transform Histogram（CEN-TRIST/CTH）are extracted respectively in the stage of feature extraction. Then the CTH feature vector is refined.Finally,HOG andsparsed CTHfeatures are combined. The experimental results show that the method has good robustness and can detect the trafficsigns quickly and accurately.