Semantic Sea-Ice Classification for Belgica Bank in Greenland

<p>Each Sentinel-1 image is tiled into patches of 256x256 pixels. The size of the images is different and we reduced to the smallest one. In total for each image are 6,400 patches [1-2, 4]. See the excel file for the 24 Sentinel-1 ids.</p> <p>The semantic classes are:</p> <ol> <li><em>Black border</em></li> <li><em>Old ice</em></li> <li><em>First-Year ice </em></li> <li><em>Glaciers </em></li> <li><em>Icebergs</em></li> <li><em>Mountains </em></li> <li><em>Young ice</em></li> <li><em>Water group</em></li> </ol> <p>The last class combines the <em>Floating ice, Water body, Water ice current and melted snow</em> defined in [3] because they have very similar physical properties.</p> <p>&nbsp;</p> <p><strong>References:</strong></p> <p>1. C.O. Dumitru, G. Schwarz, C. Karmakar, and M. Datcu, &ldquo;Machine Learning-Based Paradigm for Boosting the Semantic Annotation of EO Images&rdquo;, IGARSS, Belgium, July 2021, pp. 1-4.</p> <p>2. C. Karmakar, C.O. Dumitru, and M. Datcu, &ldquo;Explainable AI for SAR Image Time Series: Knowledge Extraction for Polar Areas&rdquo;, MDPI Remote Sensing Journal, 2021, pp. 1-21 (under review).</p> <p>3. C.O. Dumitru, V. Andrei, G. Schwarz, and M. Datcu, &ldquo;Machine Learning for Sea Ice Monitoring from Satellites&rdquo;, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-2/W16, pp. 83-89, 2019.</p> <p>4. C. Karmakar, C.O. Dumitru, G. Schwarz, and M. Datcu, &ldquo;<em>Feature-Free Explainable Data Mining in SAR Images Using Latent Dirichlet Allocation</em>&rdquo;, IEEE JSTARS, vol. 14, pp. 676-689, 2021.</p>