Iterative Spectral Index Ratio Exploration for Object-based Image Analysis of Antarctic Coastal Oasis Using High Resolution Satellite Remote Sensing Data

We introduce a new semi-automated approach by coupling spectral index ratios (SIR) and object-based image analysis (OBIA) to classify very high resolution WorldView 2 (WV 2) satellite image to extract land cover features using eCognition© software. This study aims to develop rule sets for object-based classification of WV-2 image to accurately delineate the boundaries of land cover features in the Larsemann Hills, eastern Antarctica. Multi-level segmentation process was applied to WV-2 image to generate different sizes of image objects representing different land cover features with respect to scale parameter. Several new SIRs were investigated and applied to objects at different segmentation levels to accurately classify land cover classes of landmass, man-made features, snow/ice, and water bodies. A specific attention was given to water body class to identify water areas in image level considering their different appearance on landmass and ice. The results illustrate that synergetic usage of newly developed SIRs and image segmentation in the OBIA can provide highly accurate identification of land cover classes with an overall classification accuracy of ≈98%. Furthermore, our results suggest that OBIA can contribute to powerful automatic and semiautomatic analysis for most remote sensing applications in coastal Antarctica, and synergetic use of pixel-based SIRs explores the rich geo-information of coastal areas.