A novel YOLOv11-Driven deep learning algorithm for UAV multispectral oil spill detection in Inland lakes
Abstract Lake oil spills are challenging to detect accurately due to complex oil–water interactions resulting from water flow disturbances, vegetation occlusion, and the diffusion behavior of oil films. Traditional remote sensing methods often fail to provide rapid and precise monitoring under these...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-07-01
|
| Series: | Journal of King Saud University: Computer and Information Sciences |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s44443-025-00117-z |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Lake oil spills are challenging to detect accurately due to complex oil–water interactions resulting from water flow disturbances, vegetation occlusion, and the diffusion behavior of oil films. Traditional remote sensing methods often fail to provide rapid and precise monitoring under these conditions. To address these challenges, we propose YOLO-ADHF-SimAM, a novel oil spill detection model built upon the YOLOv11 architecture. Our model integrates the self-developed ADHF module—which fuses multi-scale features using an adaptive diffusion-based hierarchical feature aggregation strategy—with the SimAM attention module to enhance key feature extraction. This integrated approach is specifically designed to capture the unique spectral and spatial characteristics of oil–water mixtures. A UAV was deployed to acquire 623 multispectral images from oil spill sites in the Daqing oilfield, forming a comprehensive dataset for model training and evaluation. Experimental results show that, compared to the baseline YOLOv11 model, YOLO-ADHF-SimAM achieves a 1.8% improvement in detection accuracy, a 6% increase in recall, a 3.3% boost in mAP@50, and a 2% enhancement in mAP@50–95. These improvements underscore the robustness and precision of our algorithm, highlighting its potential as an efficient, real-time solution for environmental monitoring and emergency response in complex inland water scenarios. |
|---|---|
| ISSN: | 1319-1578 2213-1248 |