Marine Safety Enhancement through Oil Spill Detection Using U-Net Architecture

Oil spills pose a significant danger to marine ecosystems, coastal inhabitants, and various economic activities. Early identification is crucial, as the rates of spread and impact increase, making containment more challenging. This research investigates the application of the U-Net architecture for oil spill detection. The component called skip connections, along with the U-Net encoder-decoder structure, enables precise location and segmentation of oil spills, which are typically complex and have irregular shapes, from SAR imagery. Other approaches rely on manual analysis, which is time-consuming and relatively inaccurate. By applying U-Net as a deep learning method, our approach achieved 96% accuracy in identifying oil-contaminated areas, proving to be more efficient compared to traditional methods. This increase in efficacy enhances countermeasures, response times, and resource deployment due to the technology's precise detection capabilities, ultimately benefiting marine conservation efforts.