Assessment of pixel-based classification (ARTMAP fuzzy Neural Networks and Decision Tree) and Object-Oriented methods for land use mapping (Case study: Meymeh, Ilam province)

Land use mapping is the basic tools for administrators and land planners. Severd methods have been proposed for land-use mapping. The latest and most important methods is using remotey sensed data for Land-use mapping. The aim of this study was performance evaluation of the pixel-based classification. (Fuzzy ARTMAP neural network and decision tree Methods) and object-oriented classification methods and using Landsat 8 image of 2013 for land-use mapping of arid and sem-iarid regions of Meimeh Ilam. Different land use classes were difined using training samples comperison of classification results of three different methods of fuzzy ARTMAP neural network, Gini decision tree and Object-oriented Show that the object-oriented approach, has overall accuracy of 95.30 and Kappa coefficient of 90.88 , and Gini tree decision and Fuzzy ARTMAP Neural Network methods has overall accuracy of 80.32 and 72.20 and Kappa coefficient of 68.75 and 36.18, respectively thus, object-oriented classification method having a difference in overall accuracy 14.98% and 23.1% and Kappa coefficient of 22.13% and 54.7% has a higher accuracy compared with the Gini decision tree and fuzzy ARTMAP neural network. Map area defined by the three methods of classification, are similar in farmlands, poor rangeland, and urban area. The greatest differences were observed in area of medium rangeland and minimum differences were related to the urban area.