comparing accuracy of artificial neural network, support vector machine and maximum likelihood algorithms for land use classification (case study: dashat abbas arid region, ilam province)

land use maps are the most essential information in the hand of natural resources managers.in the present research, to prepare dahat abbas land use map, landsat etm+ (1386) data was used. the image was geometrically corrected with root mean square error of less than 0.47. for classification, artificial neural network, support vector machine and maximum likelihood methods were used.finally, theland covermapwas created containing fourclasses ofagricultural lands, poor rangeland, barelandsandsand dunes. to evaluate theaccuracy ofclassification, the produced map was comparedwitha ground truthmapthroughgpsand field work. the resultsshowed thatann methodwith overall accuracyof 96.5% and a kappa of 95.16% was better than thesupport vectormachineandmaximum likelihoodmethodsinmappingland cover.this studyshowed the suitability of the neuralnetworkclassification method inprovidingland cover map of the area withhighaccuracy.