Related Experiment Video
Updated: Jan 12, 2026
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
A deep learning-based method for marine oil spill detection and its application in UAV imagery
Luhao He1, Yongzhang Zhou1, Huanrong Yang2
1Sun Yat-sen University, Center for Earth Environment and Earth Resources, Zhuhai, 519000, Guangdong Province, China; Sun Yat-sen University, School of Earth Sciences and Engineering, Zhuhai, 519000, Guangdong Province, China; Key Laboratory of Geological Processes and Mineral Resources Exploration, Zhuhai, 519000, Guangdong Province, China; Sun Yat-sen University, Institute of Carbon Neutrality and Green Development, Zhuhai, 519000, Guangdong Province, China.
This study introduces a YOLOv12 framework for detecting marine oil spills using Unmanned Aerial Vehicle (UAV) imagery. The advanced model improves early warning and response to pollution incidents, safeguarding marine ecosystems and economies.
Area of Science:
- Environmental Science
- Remote Sensing
- Artificial Intelligence
Background:
- Marine oil spills pose significant threats to ecosystems and economies due to increased resource exploitation.
- Traditional detection methods lack efficiency and accuracy, necessitating advanced solutions for early warning and response.
- Ecological impacts include habitat degradation, mass mortality, and trophic cascade disruptions, underscoring the need for rapid detection.
Purpose of the Study:
- To develop an efficient and intelligent oil spill detection framework using Unmanned Aerial Vehicle (UAV) imagery.
- To enhance the accuracy of oil spill detection and the precision of boundary delineation for marine ecological protection.
- To provide technical support for marine pollution monitoring and response efforts.
Main Methods:
- A YOLOv12-based oil spill detection framework was developed for UAV imagery.
- A comprehensive remote sensing image dataset was constructed, featuring diverse spill morphologies and marine conditions.
- The model training incorporated high-resolution input, pretrained weights, and cosine annealing learning rate scheduling.
Main Results:
- The YOLOv12 framework achieved an F1 Score of 0.7929 and mAP@0.5 of 0.8334 on the test set.
- The model demonstrated accurate oil spill detection and precise boundary regression, with smooth Precision-Recall curves.
- Stable detection capabilities were confirmed in static images and dynamic video scenarios, including target tracking.
Conclusions:
- The YOLOv12 framework offers a robust solution for intelligent oil spill detection from UAV imagery.
- This technology provides foundational support for ecological risk assessment, pollution monitoring, and wildlife protection.
- The study highlights the promising potential of the developed framework for real-world marine pollution management applications.
