Object-Based Analysis and ASTER/Terra Data for Classifying Relief Units

Abstract

The aim of this research is to develop an object-based methodology for semi-automatic mapping of relief units. The study area is the municipality of São José dos Campos, located in São Paulo State, Brazil. For this end, a multilevel hierarchical semantic network (meant for the storage and reproduction of expert knowledge) together with fuzzy logic and a set of multispectral, geomorphometric and textural variables were used. The geomorphometric and textural variables were extracted from a DEM, obtained from a stereo pair of ASTER/Terra images. The DEM altimetric accuracy was evaluated. The validation of the final relief units map was accomplished by means of object-based statistical indices derived from an error matrix, obtained by means of a comparison between the classified scene and a reference map. This reference map was produced from a visual interpretation of the stereoscopic images. Based on the obtained results, the object-based approach has shown to be a suitable method for relief units mapping.