Target Decomposition Theory In Oil Spill Detection From Sar Data

Marine oil spills are detectable using synthetic aperture radar (SAR) sensors because of the effect of oil dampening on short gravity and capillary waves. In this paper, the potential of fully polarimetric SAR data for detecting oil spills is investigated using polarimetric decompositions based on a support vector machine (SVM) classifier. First, different combinations of power and magnitude measurements of horizontal (HH) and vertical (VV) polarisations are classified using the SVM classifier, and the best combination is determined. In another investigation, the target decomposition methods, including Krogager, Freeman, Yamaguchi, van Zyl, Touzi and Holm, are assessed to detect oil spills. For this purpose, the decomposition features are computed and classified using the SVM classifier. Experiments are conducted on fully polarimetric Advanced Land Observing Satellite data. Evaluation of the results obtained indicates that the VV polarisation and the power measurement are more appropriate. Among the target decomposition methods, the Krogager decomposition method has the best result, with a 97.3% overall accuracy. According to the results, the proposed algorithm has a great capability to identify the accurate boundary of oil spills.