Efficient Algorithms for Monitoring Polar Areas Using Satellite Images

• Main Authors: Dumitru, Corneliu Octavian, Schwarz, Gottfried, Datcu, Mihai, Everett, Alistar, Hughes, Nick, Koubarakis, Manolis
• Format: Conference Object
• Language: unknown
• Published: 2020
• Subjects: EO Data Science; Arctic; Iceberg*; Polar Science
• Online Access: https://elib.dlr.de/138117/; http://eo4polar.esa.int/

When we want to classify different ice cover types in thematic maps based on satellite images, we can directly start with the publicly available image products of the European Sentinel-1 and Sentinel-2 missions and include some appropriate classification algorithm. In our case, we are mainly interested in cartographic ice maps generated from selected single images as well as time series of overlapping ice cover images of the Arctic region that can be derived either from polarized synthetic aperture radar images (in case of Sentinel-1) or from multispectral optical images (in case of Sentinel-2 that includes infrared bands). Therefore, in the following, we will concentrate on the selection of appropriate classification algorithms, where the goal of our ice cover analysis is the routine generation of multi-class thematic maps that can be used for the analysis of snow and ice cover phenomena, and for the monitoring of shipping routes. In both cases, we have to interpret large-scale target areas with discernible brightness levels and neighborhood contrast for cold (fresh) or melting (old) ice, snow cover, seawater and local water leads, ships, coastlines, and icebergs.