MULTISOURCE CLASSIFICATION OF COMPLEX RURAL-AREAS BY STATISTICAL AND NEURAL-NETWORK APPROACHES

The automatic generation of land-cover inventories by using remote-sen sing data is a very difficult task when complex rural areas are involv ed. The main difficulties are related to the characterization of such spectrally complex and heterogeneous environments and to the choice of an effective classification approach. In this paper, the usefulness o f spectral (Landsat-5 Thematic Mapper images), texture (grey-level coo ccurrence matrix statistics), and ancillary (terrain elevation, slope, and aspect) data to characterize two complex rural areas in central I taly is quantitatively demonstrated. A statistical and a neural-networ k classification approach are applied to such a multisource data set, and their classification performances are assessed and compared. The c lassification performances of the two approaches are quantitatively ev aluated in terms of global and conditional Kappa accuracies. The Zeta statistics is used to evaluate the statistical significance of the dif ferent classification accuracies obtained by the two approaches by usi ng multisource data.