A Real Time Oil Spill Detection Using Deep Learning

The increasing frequency of maritime oil spills poses significant threats to marine ecosystems, necessitating efficient monitoring and detection methods. This project presents a deep learning-based approach for identifying oil spills in satellite imagery. Using the YOLOv9 algorithm, the system classifies satellite images into two categories: "Oil Spill" and "Look-alike," enabling accurate differentiation between true spills and visually similar phenomena. The model leverages advanced convolutional architectures to detect and localize spill regions with high precision, reducing false positives commonly caused by oceanic features such as waves, debris, or algae. The proposed methodology provides a robust solution for early identification and assessment of oil spills, supporting environmental protection agencies and decision-makers in managing marine contamination events. Experimental results demonstrate the model's effectiveness in detecting oil spills across diverse maritime conditions, highlighting its potential to enhance monitoring strategies and mitigate ecological damage.

Keywords: Oil Spill Detection, Satellite Imagery, Deep Learning, YOLOv9, Image Classification, Marine Environment, Environmental Monitoring, Convolutional Neural Networks, Spill Localization, Look-alike Detections.