License Plate Location And License Plate Character Recognition Research And Implementation Of The Algorithm

With the rapid development of China’s economic and technology,communications and transportation in our country has made great progress, and carownership is growing, which has greatly changed the way people live.It can make ourlives more convenient, but at the same time it also has brought enormous pressure to thecity’s traffic management. Under the circumstances,the intelligent transportation systembegain to be applied in our lives,it make full use of digital image processing, automaticcontrol and communications technology to manage and control the entire traffic andtransport situation automatically at all-weather, now it has become one of the primarymeans of resolving the traffic congestion,and the key technique-license platerecognition system has played an important role in our lives. The license platerecognition system has a broad application in road traffic monitoring, automatedhighway toll collection, intelligent park management. These specific applications caneffectively alleviate the traffic congestion, and save on the labor costs, improvemanagement efficiency.The license plate recognition algorithm consists of three parts: the license platelocation, the segmentation of license plate characters and the recognition of characters.This article mainly discusses and researches on the algorithm of license plate locationand the recognition of characters.1. The algorithm of license plate location. This paper based on the existing licenseplate images which come from multiple different scenarios, study and design a methodof license plate location:The algorithm of license plate location of based onmulti-feature and multi-method screening.It mainly based on the rich vertical edge inthe license plate and color information,and combine with a series of tight marginal pointelimination methods and the priori knowledge about license plate.Firstly, the Prewittoperator is applied to extract the vertical edges, then use the color information in thelicense plate which the fixed color match of the characters and background to removethe no-plate region interference points.Then according to the feature that the horizontaledge points in the plate region arrange close and their point number within a certain range,a large number of interference points can be removed again. After using twotightly coupled edge points elimination methods,a kind of creative edge pointsconnection method is applied to fuse the edge points. In the next step,the region labelingmethod is applied to search the connected regions in the fused binary image,thenpositioning the plate regions according to the priori information which includingshape,size of the license plate area and the information around the license plate,thefeature of vertical projection fo license plate and AdaBoost algorithm.In the end, usingthe Hough transform to correct the tilted license plate,and precisely positioning the upand down and left and right borders of license plate.2. The algorithm of recognition of license plate characters. In this paper,thedecision trees and random trees are tried to, meanwhile according to theirshortcomings,combine with Bagging and Boosting algorithm to train classifier forlicense plate characters recognization. According to the layout of characters in licenseplate,we design and train four classifiers.they are the Chinese characters classifier thatrecognize the first character, the capital letter classifier that identify the secondcharacter,the number and capital letter mixed classifier that identify the third to the sixthcharacters,and the number,capital letter and special Chinese characters mixed classifierthat identify the seventh character.they include all kinds of characters in license plate.This paper’s algorithm all has been simulated in MATLAB,and use a lot of imagesthat taken in several different places to test and obtain a satisfactory result.