Fuzzy Recognition Of Vehicle's License Plate Based On BP Neural Networks

The License Plate Recognition(LPR) system becomes an important part of Intelligent Traffic System(ITS) with the rapid development of computer science and digital image processing . The complete LPR system includes the image acquirement the image processing, the fuzzy recognition etc. The processes of image processing and license plate Location and character recognition and Optimization of the result is called fuzzy recognition of vehicle's license plate .The essential technology of the fuzzy recognition of vehicle's license plate system lies in the digital image pretreatment technology, the plate location technology and the plate character fuzzy recognition technology. The system applies the computer image processing technology, the license location technology, the character division technology, the license character recognition technology based on BP networks. By determining the links and processes of the fuzzy identification , screening methods and algorithms according to the demands of each link function, making a comparative analysis of the key links and algorithms by experiments , the related algorithms is improved and the parameters is determined.In this paper the research which carries on to essential technology includes the image processing technology, the plate location technology, the character division technology, feature extraction of the LPR system and the optimized research on the character recognition . The image processing technology research includes: gradation, disappearing chirp, two values, image segment, tilt correction, normalization. Using HSV color space and color combination of texture analysis methods for locating the plates are fully utilized to provide the license plate image color Information ,and a five level gray image is constructed to locat the plate. And by using vertical projection and full use of the plates prior knowledge , segmented characters are got accurately.Analysis of the various characters feature extraction algorithm is carried on to compare their advantages and disadvantages, so a coarse grid feature is extracted and the char image is normalized.. The neural network and related theories are discussed, which focuses on analysis of the neural networks in Character Recognition application, the design of BP neural network classifiers and Extraction of the license plate characters .All the characteristics of the char image is input to BP network and the output of BP network is coded by binary code. Hidden layer nodes is identified by the initial use of hidden nodes in empirical formula. All of this greatly reduce the number of iterations and the size of the network and enhance the speed of the LPR.This paper introduces software design method of the license plate recognition system and describes some operation examples. The experimental result shows that system can loacate, segment and identify plates accurately.Because of the influence of practical condition ,there are many negative factors against recognition,which in turn set a high requirement for image processing and recognition algorithm.So a more adaptable and faster recognition algorithm will directly enhance the license plate recognition system performance level.