LAND COVER CHANGES DETECTION IN POLARIMETRIC SAR DATA USING ALGEBRA, SIMILARITY AND DISTANCE BASED METHODS

Abstract. Monitoring and surveillance changes around the world need powerful methods, so detection, visualization, and assessment of significant changes are essential for planning and management. Incorporating polarimetric SAR images due to interactions between electromagnetic waves and target and because of the high spatial resolution almost one meter can be used to study changes in the Earth's surface. Full polarized radar images comparing to single polarized radar images use amplitude and phase information of the surface in different available polarization (HH, HV, VH, and VV). This study is based on the decomposition of full polarized airborne UAVSAR images and integration of these features with algebra method involves Image Differencing (ID) and Image Ratio (IR) algorithms with the mathematical nature and distance-based method involves Canberra (CA) and Euclidean (ED) algorithms with measuring distance between corresponding vector and similarity-based method involves Taminoto (TA) and Kulczynski (KU) algorithms with dependence corresponding vector for change detecting purposes on two real PolSAR datasets. Assessment of incorporated methods is implemented using ground truth data and different criteria for evaluating such as overall accuracy (OA), area under ROC curve (AUC) and false alarms rate (FAR). The output results show that ID, IR, and CA have superiority to detect changes comparing to other implemented algorithms. Also, numerical results show that the highest performance in two datasets has OA more than 90%. In other assessment criteria, mention algorithms have low FAR and high AUC value indices to detect changes in PolSAR images.