Research On Traffic Sign Classification Algorithm Based On BOF

With the development of society,cars are becoming more and more popular in people's life.Intelligent Transportation System(ITS)has become the focus of traffic sign research with the advantage of improving the driving safety.However,the changes in illumination conditions,the degradation of the color of the traffic signs,the partial occlusion,and the interference of similar objects in complex scenes constitute the challenge factors for the detection and recognition of traffic signs.In view of the above situation,the author has studied the following two points:(1)In view of the interference of light conditions on road traffic sign detection in natural scene,a traffic sign detection algorithm based on color probability model based on RGB space is proposed.Because color probability model is not affected by illumination transformation and geometric deformation and has the characteristics of brightness and chromaticity separation,the color probability model is established,and the rough segmentation area of the traffic signs detected by the model is directly calculated in the direction of gradient.The graph features(Histogram of Oriented Gradient,HOG),and finally classify the HOG features by using the support vector machine(Support Vector Machine,SVM)classifier to determine the category of the HOG.This method has small calculation and can meet the requirement of real time.However,when using this method to detect multiple road traffic signs classification,it often appears that the segmentation threshold is difficult to control,and to some extent,it affects the classification accuracy.(2)The image retrieval algorithm(Bag Of Features,BOF)describes the image with only one high dimensional vector compared with the set of hundreds of operators that describe the image in other methods.It also simplifies the design of the algorithm to a certain extent,and can greatly weaken the negative effects of geometric deformation and illumination brightness.The simplification of the feature representation method and the dimension of the image feature descriptor greatly reduce the recognition rate in the process of classifying classifier.Therefore,in view of the classification of traffic signs based on the RGB space based color probability model traffic sign detection algorithm,this thesis attempts to embed the BOF methodinto it,thus forming a traffic sign classification algorithm based on BOF.The algorithm first extracts the SIFT(Scale-invariant feature transform,SIFT)features of the traffic sign possibility area,and then uses the visual dictionary generated by the K-means algorithm to select the correct traffic signs with the SVM technology to complete the traffic sign classification.This thesis mainly studies and analyzes traffic sign classification algorithm.Experimental results show that the algorithm in this thesis has better stability and accuracy,and basically meets the requirements of practical application.However,with the development of large data,the next step can be made by using statistical knowledge to analyze the multidimensional data in the image to improve the existing analysis methods,so that the traffic sign detection and classification is more efficient and convenient.