Abstract: The autuors present a decision tree based method for an easier and more accurate land-use classification using apparent reflectance values derived from ASTER images. The method starts with the correlation analyses among ASTER bands so that the optimal bands composite is found for best visual interpretation. On the derived image, apparent reflectance values represented, emphasis is paid to the analysis of the spatial and spectral characteristics of various land-use types. Spectral profiles of several typical land-use types are extracted. A decision tree model is established based on given thresholds of the apparent reflectance values. With the model the classification is then performed, associated with an accuracy assessment. The procedures are applied to ASTER images from Beian municipal region of Heilongjiang Province, Northeast China. The classified results show that the decision tree method facilitates the classification effectively. One shortcoming of the method is that some wrong classification may occur in areas where mixed spectra of different land-use types exist.

Key words: ASTER, decision tree, classification, apparent reflectance, spectral character