Ship Detection in Satellite Images from Radar and Optical Sensors Using YOLO
Yungyo Im, Yangwon Lee · 2024
Ship detection at sea can be accomplished in a variety of ways. In particular, satellites can provide extensive surveillance. However, satellite imagery has limited temporal coverage, so near-real-time ship detection can only be achieved by utilizing both Synthetic Aperture Radar (SAR) and optical satellite images. In this study, we propose an efficient ship detection method utilizing the You Only Look Once Version 8 (YOLOv8) model from both SAR and optical satellite images. Experiments were conducted using the lightweight YOLOv8s model to (1) compare the accuracy between individual and integrated models for SAR imagery, and (2) analyze the performance of models trained on optical satellite images. The experimental results indicate that the YOLO model demonstrates satisfactory ship detection results in both SAR and optical imagery, suggesting its effectiveness in constructing a real-time ship monitoring system.