Analysis Based On Wavelet Support Vector Machine License Plate Recognition Technology And Realization

License plate recognition is a classical research issue in pattern recognition field, which is very valuable in theory and practical application. The research in license plate recognition technology can make the following fields be widely used in intelligent transport, such as, digital image processing, machine vision, machine learning, pattern recognition techniques and so on. Least Squares Support Vector Machine is constructed on risk minimum principle, is able to solve the small sample training problem effectively, is used widely and achieves great success in pattern recognition. Considering the excellent time and frequency feature of the wavelet transform, in this thesis wavelet analysis is used to process the images, which has the unique advantages. SVD algorithm is used to extract the license characteristics and Least Squares Support Vector Machine is used to classify the received characteristics. The research has high theoretical significance and practical value.Major work and innovation are as follows:1. Research the processing method of the license plate images based on wavelet analysis. Reduce the noise and enhance the image by the method of enchancing images based on wavelet analysis. Use Sobel algorithm to get the edge of the license plate in order to verify the integrity of the characters. The license plate images were compressed by the method of wavelet analysis, in order to get high quality license plate image.2. Locate the license plate area with a complex background and extract the license plate, then process the license plate by binaryzation and segment the images we received. SVD algorithm was used to extract the license characteristics.3. Research the calculation in the license plate recognition. The previous ideas that license plate recognition based on support vector machine are referenced in this thesis. Least Squares Support Vector Machines(LS-SVM) was used to recognize the characters of vehicle license plates. Train and classify by regarding wavelet transformation coefficients as the input of Support Vector Machine. Two types of classifiers were designed and implemented, including number classifiers and character classifiers. LS-SVM was used as an experimental Tool.