We’re thrilled to announce that our partner UNIRI/FIDIT has published their latest study in a prestigious SCI journal , Ocean Engineering.
The study, titled “A Computer Vision Approach to Estimate the Localized Sea State,” explores a ground breaking method to improve maritime safety and efficiency through real-time sea state recognition.
In this research, they applied cutting-edge computer vision and deep learning techniques to analyse sea images captured by a single stationary camera on the ship bridge. Using this data, they trained a deep learning model to automatically classify sea states based on the Beaufort scale, employing advanced neural networks including ResNet-101, NASNet, MobileNet_v2, and Transformer ViT-b32. They also developed a unique large-scale dataset collected from a vessel over diverse sea conditions and fine-tuned our models using transfer learning.
Their findings suggest that this method could complement traditional sea state measurements, providing valuable data even when in-situ measurements are not feasible or buoy data lacks accuracy. This work lays the foundation for further innovations in sea state classification, promoting safer and more efficient maritime operations to help meet carbon reduction targets.
Stay tuned for more updates as we continue advancing maritime research!
Link to DOI: https://doi.org/10.1016/j.oceaneng.2024.118318
