The oil spill detection can be seen as a classification or segmentation problem. The purpose is to distinguish the pixels belonging to the oil slick from those representing the sea surface. The oil spills in sensor acquisitions from Synthetic Aperture Radar (SAR) typically appear as darker pixels: the density of the oil reduces the roughness of the sea surface, so that the incident microwave signal is mainly scattered away from the antenna. VV polarised SAR acquisitions should be preferred for the oil spill application. In fact, VV polarization gives higher radar backscattering from the sea surface, and therefore it provides more contrast when oil is floating on the sea surface. In optical imagery, the spectral signature of the investigated pixels should allow, in principle, distinguishing between sea and oil spill. It has to be noted that other ocean features reflecting either meteorological or oceanographic conditions can lead to the misinterpretation of the images. Common look-alikes in SAR images (i.e. dark structures resembling oil spills) are natural films, low wind surfaces, rain cells, shear zones, internal waves.
To perform the detection of oil spills, this research application builds on the tools provided in the SNAP toolbox, making them ready to process large datasets in a Cloud Computing environment. During the project the results of the application will be validated against historical oil spills acquired along the Portuguese and Irish coasts.
If you plan to use this application and have some doubts or want to share your problems or your results using it, you can write a comment on this topic, and ask for some feedback from the Co-ReSyF community.