Classification of Remote Sensing Image Using Improved LS-SVM

In this paper, an improved least squares support vector machines algorithm for solving remote sensing classification problems is presented. Support Vector Machines (SVM) is a potential remote sensing classification method because it is advantageous to deal with problems with high dimensions, small samples and uncertainty. The general idea of the proposed algorithm is that spectral angle mapping (SAM) algorithm is introduced in basic kernel functions, which make the kernel functions have better learning ability and generalization ability. From our simulation for solving remote sensing classification, the proposed algorithm indeed is very efficient.