Application of deep learning to polarimetric SAR classification

It is difficult to extract effective scattering features or describe data using simple statistical distribution for classification of polarimetric SAR. Deep learning is effective in generating complex data model since it use network graph to model data. In this paper, a classification scheme is proposed based on deep learning algorithm, and a hierarchical structure is used to classify data based on the deep learning classifier and polarimetric feature selection when there are a variety of land covers. The classification scheme is accessed using polarimetric SAR images acquired by RADARSAT-2 sensor over Singapore-area. Experiment results demonstrate the effectiveness of the proposed classification method.