Dual-polarimetric feature extraction and evaluation for oil spill detection: A near real time perspective

Oil spill detection using SAR imagery is well established and currently used operationally over European waters. However, adaptation of oil spill detection methodologies exploiting polarimetric features is still in research phase and until recently those properties have not been used for operational services. Proposed methodology introduces for the first time a combination of traditional and polarimetric features for object-based oil spill detection and look-alike discrimination in a Near Real Time environment. A total number of 35 feature parameters were extracted from 225 oil spill and 26 look-alikes and divided into training and validation dataset. Extracted features have been assessed and ranked according to their ability to discriminate between oil spill and `look-alike´. Extracted features are used for training and validation of a Support Vector Machine based classifier. Performance estimation was carried out for the proposed methodology on a large dataset with overall classification accuracy of 90% oil spill and 91% for look-alike. Polarimetric features such as Geometric Intensity, Co-Polarization Power Ratio, Span proven to be more discriminative than other polarimetric and traditional features.