Evaluate the performance of different algorithms of pixel- based classification in providing the landscape map (Case Study: Malekshahicity, Ilam province)

Abstract

Evaluate the performance of different algorithms of pixel- based classification in providing the landscape map (Case Study: Malekshahicity, Ilam province) Abstract: Nowadays, technology of remote sensing has allowed users to use this science to consider the changes arising from natural and human factors and to determine the amount of the variation due to its great progress. The researchers use several methods to classify images that each has better accuracy and efficiency compared to each other. The aim of this study is to compare the accuracy of the 9 methods of classification in malekshahi city with an area of 1739 km 2 . For this purpose, there were used the images of ETM + sensor of Landsat satellite in 2014. At first, there were done geometric corrections on the images. And finally, the map of classification of the algorithms of support vector machine, maximum likelihood, minimum distance to mean, multilayer Perceptron artificial Neural Network, Mahalanobis distance, spectral angle map, spectral information divergence, parallel surfaces and binary encoding (codes) were prepared. Results showed that the method of Multilayer Perceptron artificial neural network with back-propagation algorithms could obtainhighest accuracy and efficiency among different methods with Kappa coefficient and overall accuracy equal to 0/94 and 96.5, respectively. Mahalanobis distance method, minimum distance to mean method and support vector machine method were next priorities with overall accuracy equal to 91.35, 90.10 and 84.48. The study of the area of land use also showed that good results can be provided about the area of land use of region using classification method of artificial neural network due to high accuracy. The results can be used to extract land use maps of malekshahi city using Perceptron artificial neural network due to high accuracy. Keywords: Remote sensing, classification, landsat satellite, multilayer neural network, Malekshahi city Received (Gelis):  13.09.2016 -  Revised (Duzeltme):  07.11.2016 - Accepted (Kabul):  11.11.2016 Cite (Atif):  Yaghobi, S., Fathizadeh, H., 2017. Evaluate the performance of different algorithms of pixel- based classification in providing the landscape map (Case Study: Malekshahicity, Ilam province).  Journal of the Faculty of Forestry Istanbul University  67(2): xxx-xxx. DOI: 10.17099/jffiu.xxxxx