Gabor wavelet for road sign detection and recognition using a hybrid classifier

Abstract

Driver support systems (DSS) of intelligent vehicles analyze the image of road scenes captured by camera and detect the road signs. Then by recognizing the type of traffic sign, it can warn the driver. Most of them use the HIS color space for detection of road signs. But in this paper the YCbCr color space is used. This paper proposes a new method for both detection and classification of red road signs. The strategy consists of three steps. In the first step the input image has been transferred from the RGB color space to the YCbCr color space and the red pixels are extracted. Then the road sign object is detected from those that had been extracted as red objects. In the second step this road sign image must be convolved with a bank of Gabor wavelets and extract the feature vectors for classification. Finally in the third step these feature vectors are classified by a hybrid classifier that is composed of one-vs.-rest support vector machines (OVR SVMs) and naive bayes (NBs) classifier. The proposed method was implemented for classification of four classes of red road signs and achieved the accuracy of 93.1%. Moreover the proposed method is robust against the translation, rotation, and scale.