Integration of a Deep Learning Based Oil Spill Detection System into an Early Warning System for the Southeastern Mediterranean Sea

Kurzfassung

Deep learning based techniques have been applied in different environmental monitoring applications, including oil spill detection. This study aims to provide an oil spill detection and early warning system in order to take quick action for clean up operations for minimizing the environmental impact. Oil spills in the spaceborne Sentinel-1 Synthetic Aperture Radar (SAR) data are detected by a trained deep learning based object detector. The detections are then defined as binary masks by the segmentation method. Afterwards, the slick trajectory simulation is carried out. The system is running automatically on a regular basis when there are expected Sentinel-1 acquisitions. However, to avoid unnecessary clean up operations on false alarms, the system requires manual confirmation before sending a warning to the corresponding decision makers. The system performance should be evaluated with detailed analysis, but the feasibility of building such an automated oil spill detection and early warning system has been shown.