Alphabetic and digital characters recognition of license plate based on LS_SVM and binary tree multi-class classification

Character recognition is the focus issue on automatic license plate recognition system. Alphabetic and digital recognition is an important part of the character recognition. To improve the recognition speed, this paper presents an algorithm based on least squares support vector machines (LSSVM) and binary tree multi-class classification in the recognition of alphabetic and digital characters on license plate. Firstly preprocess the segmented character of license plate images by global binarization and normalization. Secondly, extract the whole pixels of each alphabetic character and inter-line and inter-column pixels of digital character as their eigenvectors. And then, select radial basis function (RBF) as kernel function, and train several binary LS_SVM classifiers based on the structure of binary tree multi-class classification. Finally, classify and identify the alphabetic and digital character using these classifications. Experiment results show. that the average speed for recognition is 19.3/character on a personal computer which utilizes a 2.8GHz Pentium IV processor with 512MB memory. This algorithm can be used for the fast classification and recognition.