Study Of Technology On Vehicle License Plate And Logo Recognition

Vehicle recognition is one of the crucial technologies in Intelligent Transportation System(ITS).Vehicle license plate recognition and vehicle logo recognition are two important research field in vehicle recognition.The information of vehicle is consisted of plate and logo which play an important role in many scenarios,such as garage management,charging station,traffic violation and so on.It also has great economic value and practical significance.This project makes a comprehensive understanding and analysis of the theories and algorithms related to license plate recognition and vehicle logo recognition technology in recent years,and systematically expounds the difficulties of license plate recognition technology and vehicle logo recognition technology.The development and optimization for the system of plate recognition and logo recognition is completed in PC platform.The key points of this project are to study and improve the character segmentation and character recognition of license plate recognition technology,and the vehicle logo localization and recognition of vehicle logo recognition technology.The main contributions of this paper are as follow:(1)Character segmentation:a group of essential methods in picture processing are applied for preprocessing on the pictures before segmentation.And then a character segmentation method of dynamic template combined with nonzero pixels is put forward based on it.the method includes the steps as follow.Firstly,the template of plate is set according to the character of plate char which is arranged by a certain proportion.Then that template is used for sliding in the pictures of preprocessing and the width of template is adjusted dynamically.The number of nonzero pixels located at region of 7 characters is calculated in order to contain the location of the number of maximum nonzero pixels which is used as the final position of character segmentation and realize the char segmentation.At the end,a test is made based experimental database,which reaches up to a current rate of 95.62%and takes 14.75ms on average.(2)Character recognition:the character recognition which based on the support vector machine combined with local binary pattern character is realized.That method has better accuracy ratio and less time.Some faults are happened due to the poor quality of pictures(such as incline,incomplete vague and so on)when make a statistical analysis.Therefore,this paper collects many poor quality pictures as sample and adds them to the training set of original char in order to get classifier by training them again.The accuracy ratio is up to 97.73%from previous 94.80%.The experimental result indicate that the degree of coverage by the training set has great effect on classifier.(3)Vehicle logo localization:the localization of logo is realized based on the technology of license plate localization.This method restrains the horizontal or vertical texture in order to highlight the region of logo firstly.Then it eliminates the noise around the logo.Elements of matrix structure are used for realizing the accuracy positioning.A test is made at experimental database and the accuracy rate of logo localization including plate positioning is up to 94.19%.(4)Vehicle logo recognition:the logo recognition is realized based on the SVM combined with histogram of oriented gradient.A test is made at experimental database for vehicle logo and the accuracy rate is up to 97.74%.It takes only 3.60ms and also meets the time requirement.