Study And Realization Of License Plate Reconition System Based On Wavelet Analysis

License Plate Recognition(LPR) technology License Plate Recognition as an important tool of automated traffic management and an important part of vehicle detection system plays an important role in traffic surveillance and control. Wavelet analysis is a local transformation of time and frequency, which can effectively extract information from signals, can do multiscale analysis of function or signal through the expansion and translation functions. As a result of good characteristics of time-frequency analysis, wavelet analysis has become an effective tool in signal processing and image analysis system. This paper discusses the application of wavelet analysis in License Plate Recognition Systems and other related key technology of License Plate Recognition System. License Plate Recognition system is mainly carried out in three stages: license plate location, segmentation plate into characters and character recognition.License plate location (LPL) is one of the most important roles in the license plate recognition (LPR)system.The paper projects the conception of non-orthogonal haar wavelet transform based on haar wavelet that improves the uncontinuity of haar wavelet transformation. Based on non-orthogonal haar wavelet transformation,this paper integrates the method called scanning line which is rapid and the Mathematical Morphology which can accurately find license plate,and gets one rapid algorithm of license plate Location.Experimental results indicate that the ratio of right location of license plate is higher than 95% with this algorithm.it is robust ,espetially used for detecting oblique license plate.In the stage of correction of oblique licence plate, horizontal (or vertical) edges are firstly extracted,after binarized, the image of licence plate is corrected using bilinear interpolation according to the angle of the oblique horizontal (or vertical) edge.Accurate segmentation of license plate into characters is able to lay a good foundation for the identification of characters follow-up,as is one of the key technologies in license plate recognition system..This article uses the standard projection method combined with the experience knowledge of interval to segment characters.this method overcomes the disadvantage of a simple projection method and the simple fixed interval method, and it combines the advantages of both method,so it can accurately segmentate licence plate into characters. Character Recognition which is the final step in the entire system, determines the ratio of character recognition.The quality of character feature selected determines the level of recognition, and classifier designed fundamentally determines the ratio of character recognition and efficiency of recognition.In this paper, the characteristics of characters outline and characteristics of wavelet moment are combinated as character features, then dimensionality reduction is carried out on features through principal component analysis.RBF neural network is designed to identify characters.About the above method,experiment are carried out with Visual C++ and MATLAB to validate and analyse it.The result show that overall effect is so good that recognition ratio gets to 95%.