Pixel-Based Land Classification Using Support Vector Machine and Random Forest Algorithms In Sentinel-2 Satellite Images

Abstract

In recent years, there have been rapid advances in satellite technology. In line with these developments, new and different research topics such as artificial intelligence and machine learning have emerged in the field of remote sensing. Classification methods are frequently used in collecting Earth information because wetlands such as lakes, rivers or urban areas are essential for the environment and ecological balance. It is possible to detect the current situation on the Earth with remote sensing and monitor change in time. The purpose of this study is to classify some of the agricultural lands in the city center of Konya using Sentinel-2 MSI satellite data. Pixel-based classification method and support vector machine (SVM) and random forest (RF) algorithms are used. The general accuracy of the resulting products obtained from algorithms coded in Python language was compared and interpreted.As a result of the study, the overall accuracy value obtained from the SVM algorithm was 96.7%, while the RF was 86.67% and the kappa values were 0.9535 and 0.8148 for the SVM and RF algorithms, respectively.