Baltic Sea Ice Concentration Estimation From C-Band Dual-Polarized SAR Imagery by Image Segmentation and Convolutional Neural Networks

In this study application of convolutional neural networks (CNNs) preceded by synthetic aperture radar (SAR), image segmentation for sea ice concentration (SIC) estimation over the Baltic Sea from dual-polarized C-band SAR imagery is studied. Three algorithm variants were studied and trained using FMI chart SIC or a dataset with different values generated combining pure open water blocks applying binary masks. The first two only fully ice-covered patches, based on charts, they had similar CNN structure. These differ in deriving segmentwise output. third algorithm, variant data derived as mixture patches according to charts used training. results evaluated respect an independent test dataset. very encouraging operational purposes significantly better than our earlier SAR-based algorithms. version clearly outperformed other versions SAR, reference.