Research And Achievement Of Tow-level Feature Extraction Based License Plate Character Recognition

With the development of the national economy and society, People’s living standards hasbeing improved continuously. This leads to a substantial increasing in the number of vehiclesand a lot of traffic problems. Therefore, Intelligent Transportation System (ITS) has beenadopted as an important mean of traffic management. As the kernel of ITS, license platerecognition system has importance academic significance and great economic value. Thisthesis studies the character feature extraction algorithm in the license plate recognition system.The main contents of this thesis include the following three aspects:1. For character feature extraction, three kinds of character feature including characterpixel matrix, region character edge, and structure analysis of skeleton have been studied.Especially, the principle, implementation, and advantage of this three feature extractionalgorithm have been deeply analyzed.2. A two-level feature extraction based license plate character recognition algorithm hasbeen proposed. The first level of the algorithm is a combination of the character pixel matrixand support vector machine (SVM) method, and the KNN method is applied to the regioncharacter edge feature and structure analysis feature of skeleton in the second level.3. The two level feature extraction based on license plate character recognition algorithmis designed by using the Matlab programming language.And some experiments are conductedto verify the performance of the proposed algorithm. The experimental results demonstratedthat the proposed algorithm could achieve an impressive recognition rate of the license platecharacters.