Automatic extraction of floating ice at Antarctic continental margin from remotely sensed imagery using object-based segmentation

Abstract

Information of Antarctic iceberg and sea ice are valuable to Antarctic ice melting patterns studies which are helpful to understand climate conditions and general trends of our planet. This paper presents an automatic floating ice extraction method based on image segmentation technology using region growing. It effectively solves the over-segmentation and under-segmentation problems by merging the gray, contour, location and other information of each ice-object. A pixel-based extraction method is proposed to extract the small ices within 5 pixels. LANDSAT TM data, Chinese environment disaster satellite HJ1B data, and MODIS 1B data used to detect Floating ice at Antarctic continental margin respectively. The results showed that the extraction accuracies of the three kinds of data were all higher than 81 percent, while the accuracies of both TM data and HJ1B data were higher than 90%. Object-based information extraction methods can not only obtain the total area and number of floating ice objects in the whole region, but also provide precise details of single objects, including area, perimeter, contour, average brightness. © 2011 Science China Press and Springer-Verlag Berlin Heidelberg.