Abstract—For Zhanghe areas, we utilized the Landsat5 TM
images of 2010 to complete the region's land use classification
based on the maximum likelihood method. Analyzing the
constraints of land use classification in the research area, our
classification method includes 4 aspects: cloud and hillshade were
classified at first and then reprocessed separately; water body
was classified combining supervised classification and interactive
interpretation; cities and villages were distinguished based on
prior knowledge; the slope hierarchy of arable land was
accomplished by slope of the study area based on decision tree.
67% of regions of interest (ROIs) were randomly selected for
sample training, while the rest 33% of ROIs participated in
confusion matrix validation. The experiment shows that the
classification accuracy is 91.50% and the Kappa coefficient is
0.8587.