Sea surface wind field retrieval from compressive sensing Sentinel-1 dual-polarimetric SAR images using deep learning

The sea surface wind field (SSWF) plays an important role in the marine environment and meteorological system, and is closely related to ocean movement and climate change. Spaceborne synthetic aperture radar (SAR) technology is widely used to accurately obtain high-resolution and spatio-temporal continuous information of SSWF. However, SAR high-resolution imaging faces storage and data transmission limitations, which make frequent observations impossible. The introduction of compressed sensing theory provides a solution to this problem, by effectively reducing the amount of data to improve the observation frequency. Although compressive sensing processing may lead to information loss, it is still necessary to adopt a suitable method to retrieve high-precision SSWFs from it. Current retrieval methods include traditional empirical models, neural networks, and deep learning, where deep learning methods have shown good retrieval performance. In this paper, we explore the possibility of applying deep learning to SSWF retrieval from compressive sensing SAR images, verify the effectiveness of our method, and open up new directions for future research.