Comparison of SVMs in Number Plate Recognition

Abstract

High accuracy and high speed are two key issues to consider in automatic number plate recognition (ANPR). In this paper, we construct a recognition method based on Support Vector Machines (SVMs) for ANPR. Firstly, we briefly review some knowledge of SVMs. Then, the number plate recognition algorithm is proposed. The algorithm starts from a collection of samples of characters. The characters in the number plates are divided into two kinds, namely digits and letters. Each character is recognized by an SVM, which is trained by some known samples in advance. In order to improve recognition accuracy, two approaches of SVMs are applied and compared. Experimental results based on two algorithms of SVMs are given. From the experimental results, we can make the conclusion that ‘one against one’ method based on RBF kernel is better than others such as inductive learning-based or ‘one against all’ method for automatic number plate recognition.