The Acquisition of Land Cover Information Using Three Indexes and Landsat-tm Image

To get land cover information of study city area, and according to the spectral feather of the wanted classes including water, vegetation, impervious surface, and soil at LANDSAT-TM remote sensing image, we used a method that using three indexes including normalized difference vegetation index, modify normalized difference water index, and ratio impervious surface index to extract the thematic information separately. Firstly, we use the indexes to calculate the image. Then, we choose threshold for the three thematic layers and did binarization for them too. In this step, we defined RISI according to the spectral feature and found the RISI cannot extract the impervious surface correctly, water information was also extracted. So, we calculated the RISI first, and then, used it to subtract MNDWI to get the impervious surface information and, a better result was gotten. Finally, we stack the three layers together and gave them the false color and, cause soil had very low value in all three indexes layers, the map including water, vegetation, impervious surface and soil was gotten together in one image, and it is effective to get them just by color difference. With taking the high resolution image as reference, the totally accuracy it can get was 91.3%, and the KAPPA was 0.892.