Land Cover Classification with Generated Full-Polarization SAR Data From Single-Polarization SAR Data Using Deep Convolutional Neural Network

Konferenz: EUSAR 2021 - 13th European Conference on Synthetic Aperture Radar
29.03.2021 - 01.04.2021 in online

Tagungsband: EUSAR 2021

Seiten: 4Sprache: EnglischTyp: PDF

Duan, Yan-Cui; Chen, Si-Wei (State Key Laboratory of Complex Electromagnetic Environment Effects on Electronics and Information System (CEMEE), National University of Defense Technology. Changsha, China)

With the ability to acquire full polarization information, polarimetric Synthetic Aperture Radar (PolSAR) is widely used in various applications. However, the single-polarization SAR data is more available in reality. In this paper, we present an approach of classifying land covers with generated PolSAR data from single-polarization SAR data using deep convolutional neural network (CNN). Experimental results on multi-temporal UAVSAR data show that the generated PolSAR data is visually and quantitatively close to real PolSAR data. Comparative experiments for land cover classification demonstrate that the generated PolSAR data contains more information and can improve the classification accuracy greatly.