Traffic Sign Recognition Based On Support Vector Machine

At present,traffic sign recognition is a research hotspot of unmanned vehicle and vehicle assistance system.High-quality traffic sign identification can provide real-time and accurate information of traffic conditions and traffic rules for drivers or unmanned vehicles to assist driving decision-making,so as to improve traffic safety and reduce or avoid traffic accidents.Therefore,traffic sign recognition is worth studying and full of challenges.In this paper,the Chinese traffic data set based on the intelligent processing of big data of comprehensive transportation in changsha university of science and technology is adopted for the recognition based on the support vector machine.Firstly,the original image is preprocessed: SNN algorithm is selected to filter it,Then the improved MSRCR algorithm is used to enhance the image and maintain the constancy of the color of traffic signs while increasing the image contrast.Secondly,through the study of traffic signs shape features and color features,select in RGB color space to image segmentation,according to the difference in value R?G?B range directly on the natural scene image segmentation processing.However,there are too many interferences in the segmented images.After the image is grayed,the method of combining the maximum inter-class variance method and the empirical threshold value is adopted to screen the segmented area and locate the traffic signs.Finally,in traffic sign recognition stage,the size normalization of segmented traffic signs is realized,then the multi-class classification of traffic signs based on the SVM is come true.Due to the diversity of traffic sign and correlation between shape and color,so the identification procedure is divided into two steps: coarse classification based on color-shape and fine classification based on inner region.In rough classification,traffic signs are divided into four classes according to color features and external contour : red circle,blue circle,blue square and yellow triangle.In the subclassification,the improved Hu invariant moment and affine invariant moment are combined to extract the internal information features of traffic signs so as to identify the specific meaning of traffic signs.The experimental results show that the recognition rate of traffic signcan be effectively increased to 95.89% by using above methods.