With the complicated terrain and low vegetation cover in loess plateau of North China

it is very difficult to get higher accuracy of digital image classification

only depending on the traditional spectral statistical classification method. The spatial resolution of Thematic Mapper（30m /pixel） also limit the accuracy and scale of interpreting landuse maps. In order to improve the accuracy of landuse classification in loess hilly area

a large scale relief map was used to create the digital geomorphic model （DGM）. By applying Fuzzy Mathematics theory

the DGM was used to correct the digital image classification result

so as to create the regional landuse model. The study shows that this classification improves the accuracy of landuse classification effectively

and combines the digital image classification with DTM entirely.