Performance Evaluation of Sentinel-2 and Landsat 8 OLI Data for Land Cover/Use Classification Using a Comparison between Machine Learning Algorithms

With the development of remote sensing algorithms and increased access to satellite data, generating up-to-date, accurate land use/land cover (LULC) maps has become increasingly feasible for evaluating managing changes in as created by ecosystem use. The main objective our study is evaluate performance Support Vector Machine (SVM), Artificial Neural Network (ANN), Maximum Likelihood Classification (MLC), Minimum Distance (MD), Mahalanobis (MH) compare them order generate a LULC map using data from Sentinel 2 Landsat 8 satellites. Further, we also investigate effect penalty parameter on SVM results. Our uses different kernel functions hidden layers ANN algorithms, respectively. We generated training validation datasets Google Earth images GPS prior pre-processing data. In next phase, classified algorithms. Ultimately, outcomes, used confusion matrix images. results showed that with optimal tuning parameters, classifier yielded highest overall accuracy (OA) 94%, performing better both compared other methods. addition, scenes, date was slightly more 8. parametric MD MLC provided lowest 80.85% 74.68% contrast, evaluation parameters linear 150 200 accuracies. classification increasing drastically reduces datasets, reducing zero three layers.