Efficiency Comparison of Support Vector Machine and Maximum Likelihood Algorithms for Monitoring Land Use Changes

Because the various algorithms have been developed for the land use classification by using remote sensing, the suitable algorithm selection plays an important role in achieving good results. For this purpose, by efficiency comparison of two algorithms classification i.e. support vector machines (SVM) and maximum likelihood (ML), the more precision method was determined and it was used for investigating land use changes trend. The present research was carried out using TM, ETM and OLI sensors images in Siminehroud watershed. The research results showed that SVM algorithm classified satellite images better than ML algorithm and radial basis function (RBF) kernel has the highest overall accuracy among the studied methods. Therefore, SVM algorithm with RBF kernel was used to derive land use maps and monitor land use changes in the studied periods. By analysis of land use changes trend using this kernel, it was found that during studied periods, irrigated farming from 30535ha to 67210ha, dry farming from 79909ha to 123387ha, residential from 474ha to 1934ha land uses have been increased but rangeland from 259811ha to 178397ha and water resources from 30535ha to 67210ha land uses are decreasing.