Image classification via support vector machine

With the rapid growth of images information, how to classify the images has been a main problem, and most of researchers are concerning on the neural networks to realize the images classification. However, the neural networks can not escape from its own limitations including the local optimum or the dependence on the input sample data. In this paper, another new algorithm named support vector machine, whose main idea is to build a hyperplane as the decision surface, is introduced to solve the problems. In the theory part, in order to solve the optimal hyperplane for the separable patterns problem, the method of Lagrange multiplier is transformed into its dual problem. In the application section, where it proves that the support vector machine can solve the problem of classification perfectly, with regard to the input data, the eigenvalues of the images' gray information which are treated by the method of Principal Component Analysis are abstracted as input sample. It is found that the precision of the classification could arrive at 89.66%, which is far higher than the neural networks' 41.38%.