Do Not Stop Toll Collection System In The License Plate Recognition Technology Study

The ETC(electronic toll collection)system is an important branch of the Intelligent Transportation System. It involves the callback of investment in transportation construction; and it can alleviate traffic jam in toll station. Therefore, ETC systems are being developed in many countries as the earliest application in ITS. Image Processing technologies are introduced in this paper to improve the functions of the ETC system. Key technologies such as Image Processing and Pattern Recognition are intensively studied in this paper.Number Plate Recognition is mainly studied in this paper, which includes four parts: Image Preprocessing , Number Plate Location, Character Segmentation and Character Recognition. All of the orientation algorithms which were common used deal with all the data in the image, but that can't meet the need of real-time in real-time traffic flow. The orientation algorithm given in this paper can decrease operation time sharply. The segmentation methods which were common used, such as projection method and clustering method, are liable to the interference of characters' conglutination, the segmentation method proposed in this paper can solve this problem. In regard to the deformed characters of the license plate and the varied illumination and noise disturbance conditions, a character recognition method based on extended one-class SVM is put forward in this paper, which has many merits, such as highly anti-interference quality, highly recognition rate. The details in this paper are as follow:1. Image Preprocessing: some image processing algorithms employed in this paper, including grad transform, difference of image, binarization of image, Hough transform, are introduced in detail.2. Number Plate Location: Firstly, a rough location scheme based on scanning horizontally is proposed; then an accurately location scheme based on prescient knowledge and template matching is proposed.3. Character Segmentation: Firstly, a gradient Number Plate checking method based on the changing frequency of the number of jumping dots on horizontal scanning line is proposed; then a method based on horizontal scanning is proposed to locate the top and bottom boundary of the character string; and then the template of character string on Number Plate is designed based on the character strings'structural features; finally, combined with the clustering analysis and template matching method, a method named max square difference between classes based on template matching is proposed to segment characters.4. Character Recognition: The extended one-class SVM is introduced to recognize character; three groups classifiers are trained individually for number, character and Chinese character recognition. The experiment result is satisfying.A real-time license plate recognition system is designed and implemented based on these researches. This system makes a satisfactory performance and makes the running of ETC system more automated and intelligent.