The aim of this project is to develop an automated oil spill detection and early warning system off the coast of Israel. The automated detection chain includes Sentinel-1 SAR processing, a specially trained deep learning based object detector for oil spills, and segmentation of the found oil spills into binary masks. With the oil binary masks, our partners simulate the expected oil trajectory and alert the decision makers.
A total of 9768 manually labeled oil objects, collected from 5930 Sentinel-1 scenes from 2015 to 2018, were used for training and validating the object detector.
The whole system is triggered whenever there are new Sentinel-1 scenes covering the study area. From downloading Sentinel-1 data to delivering oil binary masks to the partner institute, it takes around 2–2.5 hours.