Research On Vehicle License Plate Automatic Recognition

In the Intelligent Transportation Systems (ITS) which bases on electronic information technology, License Plate Recognition (LPR) is one of the most critical technologies. The system can automatically verify, monitor vehicle or report to the police with automatic recognition for vehicle license plate. However, because License Plate Recognition system usually works in natural environment and our country's vehicle license plates include some complex Chinese characters, the recognition rate can't achieve satisfying effect in many conditions and work efficiency of the Intelligent Transportation Systems is influenced obviously.Based on digital image processing technology, computer vision technology and pattern recognition technology, license plate identification system is deeply researches and analyses. Some improvements and innovation on license plate location and characters recognition can be made in this paper.First, in the license plate location system, this paper proposes a method which makes two-level location. In primary location, using a improved blurry matching arithmetic, the defect of plentiful account in the traditional template matching arithmetic can be conquered. In precision location, based on high change frequency about 0 and 1 of license plate horizontal orientation, the accurate license plate district can be got.Second, in the characters segmentation system, based on the transcendental knowledge which comes from license plate, this paper uses a method that combines district connectivity and gray projection to conquer some difficulties which come from come unreadable characters in the traditional arithmetic. The result displays that the characters can be more exactly segmented.Finally, in the characters recognition system, this paper discusses the theory principle of the neural network and the application method in character recognition, emphatically analyses traditional Artificial Neural Network's defects and improvements measures. Aiming at character unclear phenomena, character figure means and projection method are used to extract character characteristic. Lastly improved BP neural network is used as a method to recognition the single character.As the test result, this arithmetic can accurately locate to license plate and recognize characters. And the system's performance is good. It shows that combined pretreatments and pattern recognition techniques can improve the ability of recognition.