Study Of Character Recognition In Vehicle License Plate System

With the advancement of modern information technology and sharp increase of motor vehicles, vehicle license plate recognition system got more and more attention, which has become an important research topic in information processing technology. The system is consisted of license plate location subsystem, character segmentation subsystem and character recognition subsystem. In this paper, first an overall design of vehicle license plate system is given, then on the basis of research results, key technologies about character segmentation and character recognition are further researched.In this paper we present a method to distinguish license plate type, using license plate color and character space. This method use space color information and decision tree C4.5, as well as the character intervals of license plate, to identify between civilian license plate and specific license plate. In order to improve accuracy, we also apply character segmentation iterative methods based on the feedback of plate license color. Decision of license plate type is located between license plate location and character segmentation, which improves recognition correct rate of the overall vehicle license plate recognition system.Low-quality Chinese character recognition of vehicle license plate is always a difficult problem. With the development of intelligent traffic and pattern recognition technology, the traditional approaches which based on binary images cannot meet our practical requirement. In this paper recognition approach based on gray-scale images is applied, to avoid unnecessary structural information loss in the process of binarization. We attempt to use Gabor filters as feature extraction methods, which improve recognition rate of Chinese character from the previous 75% probably to 95% above; also we attempt to apply Local Binary Patterns(LBP) into Chinese characters recognition for the first time and achieve good improvement, which not only maintains a high recognition rate but also shortens the average time-consuming. In the further, on the basic of existing LBP operator, we bring forward an improved and efficient Advanced Local Binary Patterns(ALBP) operator as feature extraction approach, which realizes time shortened further, about 1/5 of traditional methods. Experiment results prove, compared with traditional approaches, our method is robust against the recognition of low quality vehicle plate Chinese characters and it is more adaptive than conventional approaches both on preciseness and recognition speed.