Research On Identification Method Of Vehicle Logo Based On Adaboost And SVM

With the rapid development of automobile industry,vehicle logo recognition in the field of intelligent transportation,road supervision,safety tracking,car service and others,whose application is becoming more and more prominent.According to the existing methods of vehicle logo recognition with the problems of low precision and low recognition efficiency.In this thesis,the solving method which more optimized in three stages of vehicle logo recognition is proposed.Automatic recognition of vehicle logo as a focus and difficulty,most of the thesis are based on the prior knowledge and background texture feature to locate the logo,but this method has the problem of the universal.This thesis is based on the original method,a second classifier is used to select block at the beginning of the regional vehicle logo location,and the SIFT feature is used to achieve precise positioning of the vehicle logo.Finally,the compared experiments show that this method has better positioning accuracy and generality.In the expression of its feature,the advantage of HOG and LBP feature of the image is combined in this thesis,which can more fully express logo on the edge texture,shape and contour feature on the gradient direction.In order to improve the accuracy of recognition,at the same time to ensure the efficiency of the system,the PCA method is used to reduce the dimensionality of the feature fusion in the vehicle logo.At the design stage of the classifier,in this thesis,given the consideration to the characteristics of logo samples whose type and the number is unequal.vehicle logo recognition method is proposed which combined Adaboost and SVM to achieve two different classifier combination strategies.Through the public data sets GTSRB and custom logo data set of experimental results,the joint classifier can be proved that has good classification accuracy and efficiency.at the same time,the false positive rate which is caused by the deviation of the sample number can be reduced effectively.Finally,the three stages of the implementation are fused to accomplish the automatic recognition system of vehicle logo,for the common logo,including similar logo,small sample and others,automatic detection has a good recognition.