Classification of hyperspectral remote sensing image using improved LS-SVM

Support Vector Machines(SVM) is a potential hyperspectral remote sensing classification method because it is advantageous to deal with problems with high dimensions,small samples and uncertainty.Kernel functions are key part of SVM,and they are divided into local and whole types.Different kernels can produce different classification effects.In basic kernel functions, spectral angle matching method-the classical spectral matching method-is introduced and the distance measure is taken into account.SVM can deal with nonlinear problems in classification and regression easily by using kernel functions.In the paper,an SAM(Spectral Angle Mapping)algorithm that is the classical algorithm for spectral matching recognition is introduced.By comparing with Euclid distance,a distance measure based on spectral brightness and spectral vector direction(close to spectral shape)is presented.