Very high-resolution satellite data for improved land cover extraction of Larsemann Hills, Eastern Antarctica

Abstract

We compared four different image classification methods to improve the accuracy of cryospheric land cover mapping from very high-resolution WorldView-2 (WV-2) satellite images. We used four pixel-by-pixel classification methods and then integrated the classified images using a winner-takes-all (WTA) approach. The images on which we performed the classification techniques were made up of eight-band multispectral images and panchromatic WV-2 images fused using the hyperspherical color sharpening method. We used four distinctly different methods to classify the WV-2 PAN-sharpened data: a support vector machine (SVM), a maximum likelihood classifier (MXL), a neural network classifier (NNC), and a spectral angle mapper (SAM). Three classes of land cover—land mass/rocks, water/lakes, and snow/ice—were classified using identical training samples. The final thematic land cover map of Larsemann Hills, east Antarctica, was integrated using ensemble classification based on a majority voting–coupled WTA method. Results indicate that the WTA integration method and the SVM classification method were more accurate than the MXL, NNC, and SAM classification methods. The overall accuracy of the WTA method was 97.23% (96.47% with the SVM classifier) with a 0.96 kappa coefficient (0.95 with the SVM classifier). The accuracy of the other classifiers were 93.73 to 95.55% with kappa coefficients of 0.91 to 0.93. This work demonstrates the strengths of different classifiers to extract land cover information from multispectral data collected in cryospheric regions.