Feature Selection Method OF Support Vector Machine For Polarimetric SAR Landcover Classification

摘要

In order to improve feature selection method based on SVM, we use quadpolarizationrnRadarsat-2 SAR images as experimental data.rnThe first disadvantage need to be improved is feature search method, thernexhaust algorithm is used to get the most comprehensive featurerncombination. Another disadvantage need to be improved is the evaluationrncriterions for features, some testing samples are used to validate thernclassifier to evaluate the features more precisely.rnThe results indicate that, the higher accuracy is obtained by the featurernselection method developed in this paper than by the feature selectionrnmethod based on knowledge and the non-feature selection method. Thernaccuracies of these three method are 84.96%，75.74% and 65.68%.