Laplacian support vector machine for hyperspectral image classification by using manifold learning algorithms

Abstract

For hyperspectral image classification, manifold learning based graph Laplacian is proposed in the Laplacian support vector machine (LapSVM) classifier. The manifold regularization term in LapSVM constrains the smoothness of classification function on the data manifold. Since manifold learning approach is capable of exploring the manifold geometry of data, it is suitable for calculating the graph Laplacian in the regularization term. Two manifold learning methods, local tangent space alignment (LTSA) and locally linear embedding (LLE) are utilized to obtain graph Laplacian. Experimental results indicate that the LTSA and LLE based graph Laplacian produce superior classification results than heat kernel weights and binary weights based graph Laplacian in LapSVM.