Vehicle Recognition Based On Gabor Features Simplification

Intelligent transportation system is a key part of intelligent traffic management,vehicle recognition technology is an important part of intelligent transportation system.The technology has been widely used in highway payment platform,road transport management,cases solving and so on.Therefore,it is of great important to study vehicle recognition.The vehicle recognition algorithm in this paper mainly includes three modules:Gabor feature extraction,Gabor feature dimension reduction and vehicle recognition based on sparse representation classifier.The main work is as follows:1.The feature of car face images is extracted and the feature dimension is reduced.For the problem of high dimension of Gabor feature extraction,this paper used Gabor feature simplification method.This method is inspired by the idea of local binary patterns(LBP).Firstly,the multi-scale and multi-directional Gabor features of the vehicle face images are extracted,then the Gabor feature of the same scale in different directions is selectively preserved by the feature selection rule.The feature reduction method is applied to the feature reduction stage of this paper.It aims to realize the feature dimensionality and reduce the recognition time.Then,the simplified Gabor feature is reduced secondly by using the Block-PCA dimensionality reduction method.2.The sparse representation method is introduced into the vehicle recognition,the sparse representation classification algorithm(SRC)and weighted sparse representation classification algorithm(WSRC)are studied.Moreover,inspired by the idea of K nearest neighbor,a improved weighted sparse representation classification algorithm(IWSRC)is proposed.On the one hand,the similarity measurement formula in WSRC is modified,the sparsity of sparse vectors is enhanced.On the other hand,the classification results are determined by using multiple residuals,it can correct the error of some sparse coefficient.The goal of the algorithm is improving the accuracy of the classification system.In the experimental stage.Firstly,through three comparative experiments,the experiments of vehicle recognition with Gabor feature and PCA method(G-PCA),the experiments of vehicle recognition with the Gabor simplification method and PCA method(GS-PCA),the experiments of vehicle recognition with the Gabor simplification method and block PCA method(GS-BPCA),the classifier we chosen is a classifier based on sparse representation.In the experiment,compared with G-PCA,the average recognition time of GS-PCA is reduced by 25%compared with GS-PCA,GS-BPCA is improved by 2.44%on recognition rate.The experimental results show that this Gabor simplification method is feasible.Through three groups of vehicle classification experiments:SRC,WSRC,IWSRC.The.experimental results show that the recognition performance of IWSRC is better than that of SRC and WSRC,and the average recognition rate is 91.68%,which indicates that the IWSRC algorithm proposed in this paper is effectiveness and feasibility.