Segment-based land-use classification from SPOT satellite data

Spectrally heterogeneous land-use categories cannot be adequately classified from high resolution satellite data, using conventional multispectral classification techniques. In this study, built-up land is extracted based on the spectral and the spatial properties of the segments in a spectrally classified satellite image. Object structures and classification rules have been implemented in an expert system shell ― Nexpert Object. The approach has been tested using SPOT multispectral data covering an area south of Stockholm, Sweden. The classification accuracy for built-up land improved significantly after the application of a few relatively simple rules, based on segment size and on relations between segments