Abstract: Taking part of Songzi city, Hubei province as the research area, multiple feature extraction of remote sensing in research area is discussed based on remote sensing images of Landsat 5TM in 2011 and Landsat-8 in 2017, and DEM digital elevation data of ASTGTM2. Many comparative experiments of land classification based on index have found that the indices such as EBSI,MNDWI,MSAVI,and MNDBI have higher classification accuracy; the decision tree rules are generated based on the remote sensing index and DEM characteristics,the decision tree classification model is constrcted, and the land use classification results in the research area are obtained. The experimental results show that the classification accuracy of decision tree is higher than that of SVM and maximum likelihood, and classification accuracy of Landsat-8 image is higher than Landsat 5TM; some bare soil was maily used to converte into residential land and ecological land from 2011 to 2017.

Key words: remote sensing image, DEM, decision tree, land use, accuracy