高光谱影像的多核SVM分类

Multiple kernel SVM classification for hyperspectral images

以支持向量机为代表的核方法在高光谱影像处理中得到了广泛的应用。但高光谱影像的数据特点使单核学习模型的分类具有一定的局限性。提出了一种基于多核SVM的高光谱影像分类方法。该方法以线性加权求和核为多核组合方式,从简单多核学习模型的原始问题出发,通过迭代解算单个标准SVM优化问题来实现权系数的解算,最后利用一系列两类分类器组合解决多类分类问题。通过AVIRIS和PHI影像2组实验,表明了高光谱影像的多核SVM分类方法的优势。

Support vector machine is a typical kernel method, which has been widely applied in hyperspeetral im- age processing. However, because of the characteristics of hyperspectral image data, the classification based on single kernel learning model has some limitations. In this paper, a hyperspeetral image classification method based on multi- ple kernel SVM is proposed. The multiple kernel function is formed with the linear weighted combination of the single kernel functions. Then, the weights are calculated through solving the standard SVM optimization problem iteratively starting from the original problem of simple multiple kernel learning model. Finally, a series of two-class classifiers are used to achieve the multi-class classification. The experiments on the AVIRIS and PHI images were performed, and the results prove the advantage of the hyperspectral image classification method based on multiple kernel SVM.