License Plate Recognition In Complex Background Base On HSV Space And Template Matching

As the core component of Intelligent Transportation Systems,License Plate Recognition System is mainly used in many aspects,like traffic monitoring,vehicle management and so on.It includes four parts:License Plate Location,License Plate Correction,Character Segmentation,Character Recognition.This paper gives a deep research to these four parts.The main works and achievements of this paper are as follows:1.Summarizes the common license plate location algorithms.Because traditional algorithm can not effectively locate the license plate in complex scenes,this paper presents a new algorithm that based on twice color-mark in the HSV space and projection method.The algorithm improve the accuracy of license plate location.2.In the license plate tilt correction,the paper presents a algorithm to improve the traditional rotation projection.The algorithm prejudge the tilt direction or magnitude,looking for great value to determine the projection angle adjustment,The experiment results show that it's effect is good,and it can greatly reduce the time.In order to remove license plate horizontal frame,the paper design a algorithm combine with Statistical law and projection method.To remove the vertical frame,the paper design method by judge the distance between characters and edges.The experiment results show that,license plate frame and background can be effectively removed by those methods.3.To study the license plate image binarization problem,the traditional Otsu method is improved by only search part of the gray interval for binarization threshold,which accelerate the speed of calculation and to a certain extent to ensure the effect of binarization.In character segmentation,we designed a projection method combined with a license plate template method.The method is applicable for single and double line character segmentation of license plate,its segmentation effect is good and is not sensitive to noise.4.Summarizes the common methods of character recognition,the paper use the grid characteristics,structural features of strokes and stroke density characteristics as the sample characteristics,and the method of support vector machine(SVM)was adopted to get the Chinese characters,letters and number two classifiers.The experimental results show that it can effectively identify the license plate characters.