Classification of urban remote sensing image based on support vector machines

To investigate remote sensing classification approach based on support vector machines (SVMs), we classify a remotely sensed image of urban area of Kunming, China, by SVM-based classifiers with radial basis function (RBF) as kernel function. The best values of parameter gamma (gamma) of RBF and penalty parameter C are chosen carefully through training phase. Then, data are classified by the SVM-based classifiers with the best values of parameters. Producer´s accuracy and user´s accuracy are analyzed and kappa hat coefficient is computed based on an error matrix to evaluate the approach. Our study indicates that the classification approach can yield excellent classification results.