Research On Recognition Method For Chinese Characters In License Plate

As the result of the economic development, the number of vehicles and theproblems of traffic increase gradually, where intelligent transportation system isrequired urgently. License plate recognition system is an important part of intelligenttransportation system, which is widely applied in those areas requiring license plateauthentication such as violation vehicle inspection, city surveillance system, parkinglot and expressway toll-gate.License plate character contains the Chinese characters, letters and digits.However, in comparison with the letters and digits, Chinese characters have manystrokes and complex structure. Thus Chinese character recognition is the difficultyand emphasis of license plate character recognition. Recently, scholars in China putforward many techniques of license plate character recognition, but find that therecognition rate of letter and digit is higher than Chinese characters. Therefore, tostudy Chinese character recognition method in license plate further and improve therecognition rate of Chinese characters have theoretical significance and applicationvalue.Extracting Chinese character from license plate is affected not only byenvironment but by the integrity and clean degree of license plate. Therefore, manyproblems appear, such as fuzziness, fouling and deformation, which increase thedifficulty of recognition. Considering the problems above, the main work on themethod of Chinese character recognition in this thesis covers:1. Preprocessing of license plate Chinese character: Magnitude and position ofChinese characters which cut from license plate image are different each time;therefore, normalization processing is made for magnitude and position of image.Chinese characters are adjusted to32*16to ensure the same magnitude and position.2. Coarse classification of Chinese character: In order to improve therecognition rate and velocity of Chinese character, Chinese characters are divided intothree kinds according to row column and projection.3. The design of neural network recognizer: BP neural network is used toidentify Chinese character because of its stronger anti-interference ability, fault-tolerant ability and self-learning ability. Compared with template matchingmethod, neural network method has much higher recognition rate and robustness.