The Binary Tree Of Multi-Class Support Vector Machine And The Application Of It In Image Semantic Classification

The semantic gap is implementing semantic retrieval image the biggest obstacle,Image semantic classification is the key technology in image semantic data mining.Therefore, image semantic classification becomes an important research direction in thefield of Images search.It has the important research significance to image semanticclassification new technology research.Many of the established by using support vectormachine to the classification of the images．Support vector machine (SVM) is proposedby Vapnik, a kind of learning technology, is have the aid of optimization methods tosolve the new tool machine learning problem．But the support vector machine (SVM)was first put forward in the second class classification, how will the expanded to manykinds of and applied to the image semantic classification is an important part of thestudy of this article．This paper describes image mining based on support vector machine (SVM) manykinds of classification algorithm is studied, in examining more than existingclassification after SVM algorithm is proposed based on binary tree are SVM basedimproved algorithm, and the improved many kinds of SVM used in image semanticmining．The main work is follows：（1）An overview on Support vector machines for multi—class problems isdiscussed．Several methods have been proposed including“one—against all”，“one-against-one”，DAGSVM，Classification method of multi—class SVM based onbinary tree，and so on．And their pluses and minuses and performances are compared．（2）Furthermore, BT-SVM for multi-class classification are discussed, Putforward a new kind of distance definition method (D-BTSVM) as a binary tree’sgeneration algorithms．（3）Image semantic classification classifier is designed, with the improvement ofthe binary tree classification algorithm more．This classifier for image semanticclassification characteristics． Finally, the experiment is implemented on Microsoft Visual C++6.0platformfor Coral Image Database and Recognition Image Database of data sets．The improvedbinary tree class SVM and several common types of SVM algorithm classificationresults are compared through the numerical experiments, Verifiing the superiority．...