The Study of POL-SAR Image Classification Based on SVM
基于SVM的POL-SAR图像分类研究

A method for POL-SAR image classification is presented,with which combines physical scattering mechanism,texture information and SVM.The test data is DLR ESAR L-band full polarized data of the area of Oberpfaffenhofen Test Site Area(DE),Germany.This area contains natural target,such as forest, field and manmade target,such as building,runway etc.Firstly,OEC decomposition is used to get the scattering features.The texture features of HH and HV channels are also obtained.SVM is used for feature selection and classification.Then Freeman decomposition is used to get other features.Repeating the test by adding the new features,good result is achieved.The different targets can be well classified.The test proves the efficiency of classification by combining scattering features and texture features.It also proves the validity by using SVM for feature selection.