Comparison of lithological units with different classification techniques using optical and radar remote sensing images

In this study, the lithological mapping of the region to the west of Antalya, including the Western Taurus Mountains, was made using optical images (ASTER), polarimetric SAR images (Sentinel-1) and digital elevation model (DEM) data from remote sensing data. It was aimed to use maximum similarity classification from classification methods, support vector machines from machine learning algorithms and artificial neural networks classifications as random forest and deep learning algorithms and to compare them with each other. The input data sets generated by combining optical and radar remote sensing data with DEM data in different combinations are classified with different classification algorithms. Within the scope of this thesis, the images of the thin sections at different angles were taken and their classification was also performed for the detailed mineral determination of the rocks in the pilot region of Kemer and its surroundings. Spectral mixtures were produced from the obtained mineral ratios. Spectral Angle Mapper (SAM) analyses were performed using the spectral signatures generated from spectral mixtures and spectral signatures of the existing spectral libraries of rocks, and then lithological classifications also produced. The results of all analyses were compared with the existing geological map and field observations and concluded by performing accuracy analysis.