Accuracy benefits of a fuzzy classifier in remote sensing data classification of snow

Abstract

The accuracy of snow mapping from satellite remote sensing is affected by several inconveniences -topography, cloud cover, patchy snow pack -which cause high scene variability and uncertainties in classification. When a hard classifier approach is adopted, the presence of mixed coverage pixels implies that the final product will have errors. Errors are strictly connected to the spatial resolution of the data, that rules the presence of mixed pixels within the scene. Theoretically, a soft classifier is not affected by such errors, but in remote sensing applications this has not been demonstrated. In this study the efficiencies of a fuzzy statistical classifier and of hard classification approaches have been compared adopting the 'Pareto Boundary of optimal solutions' as assessing method. Low and high spatial resolution satellite images acquired over two Alpine landscapes were used for the inter-comparison exercise. Results proved the soft classifier to outperform the traditional approaches.