Research on the land cover classification in Maqu County based on the TM image

In this study, 9 quantitative indices of remote sensing were established by using the spectrum information of Landsat TM image in 2006 and the topographical information of DEMs with the techniques of RS and GIS. Then, these indices were used to establish a discriminated function based on the stepwise discriminant analysis of spatial statistics in support of SPSS software. The wetland, grassland, river, bottomland, bare-rock and desert were extracted initially by this method. The results showed that classification precision of most ground objects was relatively high, except that wetland and grassland were mixed up to some extent. Then a template was used for further extraction. As grassland is the major land cover type in Maqu, this study also calculated the NDVI and on this basis estimated the vegetation coverage. The results showed that the moderate coverage grassland is the main land cover class in Maqu County which accounts for 83.44% of the total land area; Bare-rocks occupy 3.81%; Fields with rivers and wetlands are 1.95%, 9.98% respectively. Comparing with other studies, the result is more accurate by using a comprehensive method of stepwise discriminant analysis, “template-based” method and vegetation index when doing land cover classification, which has the certain promotion application value.