Research On Character Recognition In License Plate Recognition System

With the rapid development of economic globalization,people’s living standards have been gradually improved,and the way of transportation has changed dramatically.As an important part of the intelligent transportation system,the vehicle license plate recognition system has been widely used in the field of vehicle parking management,monitoring illegal vehicles and highway toll collecting system.In general,complete license plate recognition system includes four parts: image acquisition,license plate location,character segmentation and character recognition.Based on the completion of the license plate location and character segmentation,the paper investigates the license plate character recognition.First of all,in order to solve the low recognition rate of template matching method in license plate character recognition,especially the inadequacy in identifying the similar characters accurately,this paper proposes a method of license plate recognition based on template matching method combining with local HOG feature.Firstly,template matching method is used for preliminary identification of all the characters of license plate.Then the paper extracts the biggest difference of HOG feature in the similar characters of the license plate and the template to construct the feature vector.Finally,according to the Euclidean distance between the feature vector to measure the similarity of the license plate character and the template character,and then complete the second recognition.The experimental results show that this method is effective in solving the problem of false recognition of similar character and the recognition rate is significantly improved.Secondly,with the influence of different tilt angles,in order to solve the the shortcomings of template matching with local HOG feature,by combining with the advantages of LBP and HOG features,this paper proposes a method of the license plate character recognition with the fusion of local HOG and layered LBP feature.Firstly,the paper uses the template matching method for preliminary identification of all the characters of license plate.Then a small edge feature of the biggest difference in the similar character of the license plate and the template are extracted by using HOG operator,and then the layered texture feature of the same area block of HOG in the similar character of the original licenseplate and the template are extracted by using LBP operator.Next,serial feature vectors are constructed with serial fusion of the edge feature and the layered texture feature.Finally,according to the Euclidean distance between the feature vectors to measure the similarity of the license plate character and the template character,and then complete the second recognition.The feature of LBP is mainly used to extract the texture information of the image and has significant robustness to rotation.The feature of HOG is mainly used to extract the edge information of the image and has certain robustness to illumination and character segmentation.The identification performances of the 11 algorithms by comparing the experimental results show that the single LBP method and HOG method are far less effective than the method put forward in this paper.In terms of the influence of character segmentation,illumination and tilt angle,the method of this paper is quite robust.