Classification of SAR and PolSAR images using deep learning: a review

ABSTRACT Advancement in remote sensing technology and microwave sensors explores the applications of remote sensing in different fields. Microwave remote sensing encompasses its benefits of providing cloud-free, all-weather images and images of day and night. Synthetic Aperture Radar (SAR) images own this capability which promoted the use of SAR and PolSAR images in land use/land cover classification and various other applications for different purposes. A review of different polarimetric decomposition techniques for classification of different regions is introduced in the paper. The general objective of the paper is to help researchers in identifying a deep learning technique appropriate for SAR or PolSAR image classification. The architecture of deep networks which ingest new ideas in the given area of research are also analysed in this paper. Benchmark datasets used in microwave remote sensing have been discussed and classification results of those data are analysed. Discussion on experimental results on one of the benchmark datasets is also provided in the paper. The paper discusses challenges, scope and opportunities in research of SAR/PolSAR images which will be helpful to researchers diving into this area.