Character Segmentation And Recognition In Vehicle Plate Recognition System

With the rapid development of domestic economy, the number of vehicles hasincreases quickly. More and more funds are put into the construction of roads,highways, parking lots. Roads become broader and parking areas have more space fordrivers; however, as a very important part of transportation system, management doesnot follow the steps of hardware constructions, almost everything is done by humanbeings. To utilize the convenience of modern transportation, a way should be found tomake good use of the existed transportation constructions. License Plate Recognitionis one of the research subjects in which the technology of computer vision andrecognition applied. The purpose of License Plate Recognition is to realize theautomatic vehicle supervising without other special equipment. So the automaticmanagement of traffic system will be much more convenient. This system is one ofkeys in advancement in traffic system.This thesis comparatively explains the tasks and problems and does analyticresearch across all the phases of the system, and then use BP nerve network basicallycomplete the car license recognition system design and the realization.What we do in the thesis is following:1. In the thesis, we introduce some traditional method of character segmentation andrecognitions. At last we give the method of character segmentation and characterrecognition we used in this thesis.2. In the locating phase of the system, through analyzing the valleys of the smoothedhorizontal gray texture projection, we perform minor-adjusting to the bottom andtop position determined by the rough locating operation; in the phase of charactersegmentation, we improve the operation by analyzing the valleys of the smoothedprojection of the targeted pixels from left to right and by using the width, height and other basic information of the processed plate character;3. In character recognition, we choose the feature of rough grid as the feature ofcharacter, and directly input the improved unitary character originality feature toBP neural network classifier to recognize the license plate character. To improvethe recognition rate of the character recognition system for vehicle license plate,the thesis analyzed the result of test, and put forward a series of measure, forexample, designed a careful neural network classifier to get detail feature ofcharacter, which is analogical and promiscuous. The result shows that the methodcan maximally improve the solidity performance of character network.The plate recognition system in this thesis is a typical application in the area ofpattern recognition. Its general ideas and detail design can be expanded to otherapplication areas with the similar tasks of classifying, and it has a prosperous future.