Oil Spill Detection Using Sentinel-1 SAR Data

Authors : Abdurahman Yasin Yiğit, Halil İbrahim Şenol

Pages : 120-132

View : 46 | Download : 50

Publication Date : 2025-04-30

Article Type : Research Paper

Abstract :Oil spills present a substantial threat to marine ecosystems and coastal economies, necessitating efficient and accurate detection methods. Driven by the necessity for dependable monitoring in a variety of environmental conditions, this study utilizes Sentinel-1 Synthetic Aperture Radar (SAR) data and the Mask R-CNN deep learning model to detect and delineate oil spills. Sentinel-1 was selected for its capacity to acquire data irrespective of weather conditions or time of day, thereby ensuring consistent monitoring. The Mask R-CNN model was selected for its ability to perform precise, pixel-level segmentation, enabling accurate spill boundary detection. The model\\\'s performance was evaluated using a dataset comprising 381 Sentinel-1 images from diverse geographic and environmental contexts. The model demonstrated an overall accuracy of 80% for MV Wakashio and 81% for MK Princess Empress, with an Intersection over Union (IoU) of 76% and 74.8%, respectively. These results underscore the model\\\'s efficacy in discerning oil spills from false positives, such as algal blooms and sediment patterns. The proposed methodology demonstrates a clear advantage over traditional techniques and exhibits scalability for real-time applications.
Keywords : Petrol Sızıntısı Tespiti, Sentinel-1, Çevresel Koruma, Afet Yönetimi, Sentetik Açıklıklı Radar