Land cover is one of basic data layers in geographic information system for physical planning and environmental monitoring. Until recently Maximum Likelihood classifier (MLC) has been the most common method used for supervised classification of remotely sensed data. Increasingly, nonparametric classification algorithms such as decision trees are being used. Decision tree classifiers have, however, not been used as widely by the remote sensing community for land cover classification despite their nonparametric nature and their attractive properties of simplicity, flexibility and computational efficiency. The study area selected for this experiment is a protected area locates in north-east Iran. IRS-1D LISS-3 data acquired on 5th May, 2003, is used. The area includes seven types of land cover (Wood land, Open shrubland, sparse shrubland, Meadow, Bare land, Farm land and City area). The results of this study suggest that the supervised classification (MLC) producing an overall accuracy of about 72%. The Decision tree method, which improves the classification accuracy, was applied and the classification accuracy was increased by about 7 percent to 79%.

Keywords