Visualization for exploration of uncertainty related to fuzzy classification

Abstract

This study presents a new visualization tool for exploration of the results of a supervised fuzzy classification. 3D isosurfaces are introduced to visualize the shape and extent of class clusters in a 3D feature space plot. These isosurfaces are linked to an image display showing the spatial extent of classes. The dynamical link between feature space, image space, and user interaction with an uncertainty threshold allows for interactive exploration of the spatial and thematic aspects of uncertainty related to a fuzzy classification. The visualization tool is illustrated with a fuzzy classification of a Quickbird image of sub-Antarctic Macquarie Island.