Classification of Sea Ice Types in Sentinel-1 SAR Data Using Convolutional Neural Networks

Hugo Boulze,Julien Brajard,Anton Korosov

doi:10.3390/rs12132165

Hugo Boulze, Julien Brajard + Show 1 more

Open Access

https://doi.org/10.3390/rs12132165

Copy DOI

Journal: Remote Sensing	Publication Date: Jul 7, 2020
Citations: 75	License type: CC BY 4.0

Affiliation: Université Gustave Eiffel

Abstract

A new algorithm for classification of sea ice types on Sentinel-1 Synthetic Aperture Radar (SAR) data using a convolutional neural network (CNN) is presented. The CNN is trained on reference ice charts produced by human experts and compared with an existing machine learning algorithm based on texture features and random forest classifier. The CNN is trained on two datasets in 2018 and 2020 for retrieval of four classes: ice free, young ice, first-year ice and old ice. The accuracy of our classification is 90.5% for the 2018-dataset and 91.6% for the 2020-dataset. The uncertainty is a bit higher for young ice (85%/76% accuracy in 2018/2020) and first-year ice (86%/84% accuracy in 2018/2020). Our algorithm outperforms the existing random forest product for each ice type. It has also proved to be more efficient in computing time and less sensitive to the noise in SAR data. The code is publicly available.

Highlights

For secure navigation and offshore activities, sea ice concentration and type in polar regions should be monitored and forecasted with high accuracy
The main goal of our research is to develop a convolutional neural network (CNN)-based algorithm for the classification of sea ice types on dual-polarization Synthetic Aperture Radar (SAR) data from Sentinel-1
Resolution of the CNN ice chart is much higher than the manual one and captures many small details including thin filaments of ice in open water, cracks in sea ice filled with young ice or water

Summary

Introduction

For secure navigation and offshore activities, sea ice concentration and type in polar regions should be monitored and forecasted with high accuracy. Ice concentration is defined as the ratio of the pixel area covered by floating sea ice to to the total area. Ice type can be defined in terms of stage of sea ice development—from newly frozen smooth ice (Nilas), to deformed and roughened Young Ice, to thick ice cover that survived summer melts (Old Ice) with several intermediate stages [1]. SAR images are analyzed by ice analysts in operational centers for manual classification of ice types and drawing of ice charts. With the availability of more than 100 Sentinel-1 SAR images in the Arctic ocean per day this procedure understandingly requires a significant effort and human power

Objectives

Results

Discussion

Conclusion