Oil Spill Detection in SAR Imagery Using an Optimized U-Net Network

This paper addresses the automatic detection of marine oil spills using Synthetic Aperture Radar (SAR) satellite imagery, with a focus on optimizing deep learning techniques for improved segmentation performance.Given the ecological, economic, and health-related risks posed by oil spills, enhancing monitoring systems through automated approaches is a critical priority.SAR imagery offers an allweather, day-and-night observation capability, making it highly suitable for maritime surveillance.In this study, we propose an optimized U-Net architecture tailored for oil spill detection, incorporating domain-specific preprocessing and aggressive data augmentation strategies to enhance model generalization and robustness.The methodology is evaluated on a preprocessed SAR dataset, and its performance is benchmarked against state-of-the-art methods.Results demonstrate the effectiveness of the proposed approach in accurately segmenting oil-contaminated areas, contributing to the development of real-time marine pollution monitoring systems.