A Near Real-Time Automated Oil Spill Surveillance System Using SAR and its Application to a New Study Area

This study demonstrates a near real-time marine oil spill surveillance system based on spaceborne synthetic aperture radar (SAR). The deep learning-based object detection algorithm automatically targets oil spills in the SAR scenes. Afterwards, the segmentation method retrieves the exact oil coverage. The system takes about 17 min, from obtaining four scenes to delivering oil slick information to the users, and it can even be run on less powerful hardware. In addition, we evaluate the potential for global oil pollution monitoring by applying the system to the Gulf of Mexico, an area that fundamentally differs from the training region.