Özet:

The usage areas of satellite systems and remote sensing technologies are considerably expanding. It is frequently used, especially in determining land use classes and land cover changes, and the thematic maps produced can be presented as a data source. In this study, land use classes were determined and mapped using supervised classification techniques such as maximum likelihood (ML), support vector machines (SVM) linear function, SVM radial function, SVM polynomial function, SVM sigmoid function and artificial neural networks (ANN) using Landsat 8 Operational Land Imager satellite images. For this purpose, Eleman Forest Planning Unit was selected, and the stand map of this area was used as reference data. Both classification success and the McNemar test were used to compare the performance of six different supervised classification techniques. According to the results, kappa and overall accuracy values of all classification techniques were found to be 0.80 and above 80%, respectively. The highest kappa (0.8488) and overall accuracy (89.1442%) values were obtained for the SVM polynomial function. According to the McNemar test results, there was no statistical difference between the most successful SVM polynomial function and ANN method (χ2<3.8414). The SVM polynomial function and ANN approach were found to be the most effective classification methods among the employed supervised classification techniques.

Anahtar Kelimeler: