Automated Sea Ice Classification Using Sentinel-1 Imagery

Abstract

Sentinel-1A and 1B operate in Extra Wide swath dual-polarization mode over the Arctic Seas, and the two-satellite constellation provides the most frequent SAR observation of the Arctic sea ice ever. However, the use of Sentinel-1 for sea ice classification has not been popular because of relatively higher level of system noise and radiometric calibration issues. By taking advantage of our recent development on Sentinel-1 image noise correction, we suggest a fully automated SAR image-based sea ice classification scheme which can provide a potential near-real time services of sea ice charting. The denoised images are processed into texture features and a machine learning-based classifier is trained by feeding digitized ice charts. The use of ice chart rather than manually classified reference image makes enable an automated training which minimizes the effects from biased human decision. The resulting classifier was tested over the Fram Strait area for an extensive dataset of Sentinel-1 constellation acquired from October 2017 to May 2018. The classification results are shown in comparison with the ice charts, and the feasibility of the ice chart-feeded automated classifier is discussed.