Sar İmgelerinde Gözetimsiz Sınıflandırma Yöntemleri ile Arazi Örtüsü Sınıflandırması
View/ Open
Date
2019-09-26Author
Yumuş, Duygu
xmlui.mirage2.itemSummaryView.MetaData
Show full item recordAbstract
The land cover term refers to the natural and man-made physical cover of the Earth. It is also defined as the biophysical state of the Earth's surface, such as topography, surface water and groundwater, soil and human structures. Land cover information contributes to the research of important environmental issues such as the change of forested areas on the Earth by years, determination of increase and decrease in marine pollution, detection of contradictions in the Earth, damage assessment after natural disaster. Land cover information provides essential and important information on urban planning and management, resource consumption, urban development, and determination of year-to-year change in residential areas.
Synthetic Aperture Radar (SAR) systems are an important source of information for the coverage of land cover by RF waves. The SAR systems operate by transmitting RF waves to the ground surface by a radar transmitting unit and sensing and processing the scattered waves from the surface by the radar receiver unit. By analyzing and classifying SAR images, land use information of a region including land use, land cover, various land statistics and indicators can be obtained. C-Band Synthetic Aperture Radar (SAR) images of the SENTINEL-1 satellite and X-Band AFRL SAR images of the Air Force Research Laboratory (AFRL) were used in this study. Synthetic Aperture Radar is an active remote sensing system that generates its own radiation. With this feature, it is not affected by changes in weather events and can provide images in all weather conditions.
In this thesis, the land cover of SAR images was with different unsupervised classification methods and examined extensively. Moment Based methods, Principle Component Analysis (PCA), Eigenface, Kernel PCA and Autoencoder feature extraction methods have been studied on unsupervised classification of different terrain types in SAR images. Computer simulations of algorithms were performed and comparative studies were conducted.