The Research About License Plate Recognition

With the economic and social development and increasing level of people living standardday by day，the demand for transportation is increasing, while intelligent transportationsystem (ITS) plays a very important role in the traffic regulation and control as well as itsmanagement.The key technology used in license plates recognition system includes image processingtechnology, license plate location technology, character segment technology and characterrecognition technology. In image processing, original vehicle image turns into the gray imagefirstly which then will be enhanced by gray transformation, histogram equalization andmathematical morphology, and last the noise is removed by improved median filtering. At thesame time, three image binaryzation methods are listed, and OSTU method of globalthreshold value is used in this paper.The prior knowledge of the license plate combining with projection was used in licenseplate location. The prior knowledge mainly made use of geometric features of the normallicense plate, such as the ratio of height and width, area and gray jumping regularity existedbetween license plate character and the grounding; the projection method used the method ofhorizontal projection and vertical projection.In the license plate dismemberment, correct theslantwise ones by Hough transformation which bases on that duality property between pointand line, which is used to look for a straight line with maximum length firstly, and thencombining with conglutination and discontinuous division of characters, compart charactersby elevation view. Character is the core of the whole system.Before character recognition, the isolated single characters will be normalized operationto extract characters standard feature which is convenient for character recognition. Characternormalization introduced here contains normalization of position, size and thickness of stroke,this paper is mainly introduces the normalization of size. In character recognition, BP neuralnetwork and its main ideas and algorithm is introduced in detail, and take the extractedcharacter’s features into neural network to training after it is normalized. After study of the neural network, identify license plate character which will be identified.This paper gives some improvements for the4process of vehicle license platesrecognition system which on the basis of the study of existing research results, and thoseimprovements also achieve good effect.