Oil slicks, look-alikes and other remarkable SAR signatures in Sentinel-1 imagery in the Eastern Mediterranean Sea in 2019

Kurzfassung

Spaceborne synthetic aperture radar (SAR) has been applied to detect oil slicks. Previous studies applied machine learning and deep learning techniques to automate oil slick detection. However, a large amount of data is generally required to train such a model. It highlights the importance and necessity of a publicly available oil slick dataset. This dataset provides annotations of oil slicks located in longitude from 30°E to 36°E and in latitude from 31°N to 34.7°N in the Eastern Mediterranean Sea, observed from Sentinel-1 SAR in 2019. The annotation and inspection of the oil slicks were initially done in the framework of a previous study by the authors (Yang, Y.-J. et al., 2024). On top of that, images with oceanic and other phenomena, which can also manifest similar SAR signatures and are considered look-alikes, are also included in the dataset as a no-oil set. A well-developed oil spill detection system should be able to not only detect oil spills but also avoid the detection of these look-alikes. These look-alikes were collected on a larger extent, ranging between 27.1212703°E and 36.0881997°E in longitudes and 29.2991798°N and 36.3715771°N in latitudes. There are, in total, 1365 image patches with 3225 oil objects in the oil set and 2290 image patches in the no-oil set.