The Research Of Vehicle License Plate Recognition Technology Based On SVM And Neural Network

License Plate Recognition System (License Plate Recognition System, LPRS) is image processing and pattern Recognition technology in Intelligent transportation System (Intelligent Transport System, ITS) of the important applications, are widely used in the Intelligent parking lot management, highway vehicles video monitoring, traffic monitoring and electronic police, etc. However, due to the subjective reasons such as diversity of reality, environmental impact, equipment difference,etc. and other objective reasons such as license plate abrasion, cover are difficult for higher for license plate recognition technology.Therefore,based on the study of license plate recognition technology at home and abroad on the basis of the latest research results,Extraction of the related algorithm, the main technical points of the feasible.Through image processing technology, access to pretreatment of license plate,by training the SVM model to identify the real license plate, and then split out of a single license plate characters, and using artificial neural network for license plate character recognition. In this paper, the main work is as follows:(1) License plate localization aspect:is put forward based on the license plate edge character and the SVM (support vector machine) with the combination of license plate location method. The basic principle of this method:first of all, using a gaussian filter to smooth the color vehicle images; Gray, binarization, then using Sobel operator for positioning, then use morphological corrosion, expansion of computing the vehicle image, looking for license plate candidate start-point, get some candidate license plate images, finally through the trained SVM model selection of the real license plate.(2) Character segmentation:in the license plate image with gray processing, binarization processing after get a clear picture, by judging jump frequency remove upper and lower borders, tangerine,combined with contour method and plate prior knowledge to the character image segmentation. On the basis of the characteristics of image of Numbers and letters connected, the contour method to get the number of the license plate characters and alphabetic characters; License plate prior knowledge is used to solve the problems of Chinese characters is not connected.(3) Character recognition:first extracted character features, and then use OpenCV provided by neural network to recognize Chinese characters, letters and Numbers.In this paper, uing Visual Studio 2013 to this article, The experimental results show that the algorithm is accurate and recognition rate is higher, the characteristics of fast.