Supervised classification categorized the image into eight landuse classes (healthy tea, moderately affected tea, affected tea, river, riverbed, settlements, scrubs and barren land) as shown in Figure. 17. More classes of vegetation types were avoided to minimize misclassification and confusion among the classes. The classified images were validated with the ground truth data. Maximum Likelihood Classification (MLC) gave the best results as compared to Minimum Distance to Mean (MDM) and Parallelepiped Classification. The overall accuracy of the classified image in MLC was 87.42%.  Figure 17: Landuse/Landcover Map of Sonitpur District (LANDSAT ETM+, December 2001) |