Bakirci, Murat2025-03-172025-03-1720252352-4855https://doi.org/10.1016/j.rsma.2024.103975https://hdl.handle.net/20.500.13099/1956Ship detection from satellite imagery is a powerful tool in marine science, offering crucial understanding of vessel traffic patterns, fishing activities, and environmental impacts on marine ecosystems. The ability to monitor ship movements on a large scale aids in assessing anthropogenic pressures on sensitive habitats, enforcing regulatory compliance, and supporting conservation efforts in protected areas. This study offers a fresh perspective on ship detection in oceanic environments by pioneering the evaluation of YOLOv9, a cutting-edge detection algorithm, within this domain for the first time. Applying a diverse set of data augmentation techniques significantly improved the algorithm's ability to detect small ships. Additionally, atmospheric scattering effects commonly present in satellite images were mitigated through filtering, further enhancing detection performance. With a remarkable increase in speed and significantly superior performance, particularly in detecting small ships and minimizing detection time, YOLOv9 emerges as the premier candidate for real-time applications and timesensitive critical operations.eninfo:eu-repo/semantics/closedAccessShip detectionYOLOv9Ocean monitoringDeep learningMaritime safetySatellite imageryArticle10.1016/j.rsma.2024.10397581Q2WOS:0013925702000012-s2.0-85212577627Q1