Traffic Sign Recognition Based On HOG-CTH Fusion Features

With the development of urbanization,intelligent transportation system(ITS)has become an important means to solve the urban road traffic.Intelligent transportation system mainly includes intelligent infrastructure and intelligent vehicle,it is a comprehensive information management system,mainly set communication,detection,control and computer technology in one.As one of the important components of intelligent vehicles,road traffic sign recognition system(TSR)needs to collect and identify real-time traffic signs information in the vehicle driving process,and it can also promptly instruct or alert the driver to assist driving,it can also control the vehicle directly even in special circumstances to ensure driving safety and to prevent traffic accidents.Traffic sign recognition system mainly includes the detection of traffic signs under the natural scene and the classification and identification of traffic signs.Although the previous scholars have spent many years studying these problems,they still do not solve these problems well.The main reason is that the robustness of the algorithm in complex environment is not good.This paper mainly studies the traffic sign recognition from the aspects of traffic mark image preprocessing,segmentation,feature extraction and recognition,and it has achieved good results.The main work of this paper is as follows:1.Road traffic sign image preprocessing.In the pretreatment stage,the HSV spatial histogram equalization method is used to enhance the traffic road images collected in the natural scene,and the median filter is used to remove the noise.2.Traffic mark location based on color segmentation and shape feature in HSV space.At this stage,according to the characteristics of color space model,the HSV color space is used to roughly segment the traffic signs,and then the binary image is expanded and filled by the morphological processing.Finally,the shape features are used to accurately locate the traffic signs.3.Feature extraction and classification of road traffic sign.Because a feature often does not fully describe the characteristics of a traffic sign image,two or more features can be combined to enhance the ability to express features by complementarity between features.In this paper,the HOG-CTH fusion feature is extracted from the traffic mark image.The fusion feature is composed of the direction gradient histogram HOG feature and the statistical change histogram CENTRIST / CTH characteristic.It has the advantages of strong expression ability,excellent classification effect and low feature vector dimension.In the machine learning algorithm,the support vector machine classifier is designed to classify the traffic signs by studying the statistical theory and support vector machine SVM principle.4.Applied experimental research.Common traffic signs in the actual scene are selected as the research object.It is found that the proposed traffic sign recognition algorithm can identify the traffic signs in the natural scene and has a high recognition rate,it also meets the real-time requirements at the same time.