Use of Remote Sensing Radar Images for Offshore Oil Slick Detection in Oil and Gas Domain: Manual and Automatic Interpretation

Natural and anthropogenic hydrocarbon oil slicks can be present anywhere on the surface of the world's oceans. Oil companies are interested in detecting natural hydrocarbon slicks (Seeps) for Exploration as well as anthropogenic hydrocarbons slicks (Spills) for Environmental purposes. To meet these needs, hydrocarbon detection studies in the offshore domain are essentially carried out using Synthetic aperture radar (SAR) data on which oil slicks appear as `black regions'. Today, segmentation of oil slicks is done manually by photo interpreters who manually or semi-automatically draw the contours of the potential oil slicks.Nowadays, multiplicity of SAR sensors considerably increases the amount of available SAR data, which is an advantage as the availability of images is no longer a problem, but also an issue as the photo interpreter gets more and more data to analyze in a shorter period of time. The development of Deep Learning methods is a very promising approach in the context of image processing and pattern recognition to cope with this increasing flow of data. Several solutions are currently being studied at Total and this paper presents the first encouraging results obtained.