Fuzzy logic classification for the extraction of surface parameters in alpine areas

Abstract

In alpine areas the classification of land use from optical multispectral remote sensing data often fails using the standard maximum likelihood (ML) approach. Due to the heterogeneity of surface covers and the high variability of environmental conditions separability of spectral classes is rendered very difficult. To overcome these difficulties in the land use classification of a mountainous watershed, a fuzzy logic classification algorithm is applied. The spectral probabilities of the ML classifier are combined with the terrain properties altitude and slope, that in a fuzzy manner are responsible for the spatial distribution of landcover types. It is shown, that the incorporation of such additional fuzzy geographic information can substantially improve the results of land use classification in complex mountainous terrain.