An Improved Support Vector Machine And Its Application In Image Segmentation

The method of Support vector machine, which is based on statistical theory, isconsidered as a new machine learning method; and It’s also to be a good instead of thetraditional machine learning method.This method can also performs many the uniqueadvantages in the Small Sample, Nonlinear and High Dimensional Pattern Recognition.At present, it is mainly used for binary classification problems, and not widely used tomulti-class classification, and its applying to multi-class classification is still a problemwhich was worth to be studied. The image segmentation methods which were based onthe traditional learning theory, such as neural network method was widely used,however there is poor generalization, network structure is difficult to determine, proneto learning and less learning problem.Recently,people has great interested in.themethod of using the SVM to segment image.From the support vector machine and the present situation of image segmentation,This paper introduces several important concepts, such as statistical learning theory,the basic theory support vector machine, multi-class support vector machine, and theexisting various multi class classification support vector machine was also introduced.By comparing the existing multi-class support vector，several distance is defined inthis paper, An improved support vector machine which was based on" one-against-one"support vector machine was defined, the improved support vector machine is mainly toimprove the distance and speed testing as its purpose and we also use the imagesegmentation experiment to suggest the advantage of the improved support vectormachine approach in image segmentation.