基于支持向量机的多目标分类和识别

摘要

Support Vector Machine(SVM),which was developed for binary classification initially, now is widely used in many research fields like pattern recognition. However,its application to multi-classification is inefficient. This paper did research on the extension algorithms of SVM from binary classification to multi-classification,and found that Directed Acyclic Graph(DAG-SVM)is one of the most popularly used. Therefore,this paper focuses on its application of multi-classification in military area,and achieves recognition of many military objects,such as soldiers,armored cars,low altitude targets,and so on.%支持向量机（SVM）算法广泛应用于模式识别等领域，但是SVM最初是针对二类别分类提出，在多分类识别中稍显逊色。对将SVM由二分类扩展到多分类的算法进行了研究，发现有向无环图（DAG-SVM）是其中用的最多的算法之一。因此，针对军事领域图像的多目标分类，选择有向无环图算法来实现军事图像中单兵、装甲、低空等多目标的分类识别。