Land Cover Mapping in Zagros forests using satellite images (Case study: Northen Ilam Forests)

Land use classification using remotely sensed images is one of the most common applications in remote sensing, and many algorithms have been developed and applied for this purpose in the literature. The Objective of this research is to assess the ability of three algorithms, Support Vector Machine (SVM), Classification Tree (DT) and Fuzzy Artmap (FA) for forest land mapping in Northern forests of Ilam province using LissIII dataset of the year 2007. These algorithms are a group of supervised classification algorithms thathave been recently used in remote sensing. Geometric and radiometric corrections were performed on LissIII dataset. Then, through field surveys, the land cover classes were defined and training areas were selected. The best spectral bands were selected using class signatures and the transformed divergence severability index. Supervised classification of classes was accomplished using SVM, DT and FA classifiers. The results were assessed using sample ground truth map through systematic networks of 550 meter distances. The results showed that forest lands were more easily separated from the other land cover classes. While comparing the overall accuracy and kappa coefficient for the three classifiers, results showed that the Support Vector Machine classifier exhibited the highest overall accuracy (94%) and kappa coefficient (91%).