Study on Land Classification of PolSAR Data by Using Support Vector Machine

Polarimetric Synthetic Aperture Radar (PolSAR) has been attracting attention in ground target detection and classification. Various methods, such as model-based scattering power decomposition, eigenvalue analysis and so forth, have been proposed for the analysis of PolSAR data. Recently, classification techniques by using machine learning have been intensively studied to reduce misclassification. In this paper, we present some experimental results of land classification of ALOS-2/PALSAR-2 data by using support vector machine. We confirm in this report that support vector machine analysis is more accurate than the conventional unsupervised classifier such as model-based decomposition.