Improving Marine Ecosystem Monitoring Through Computer Vision

Key Takeaways

Researchers created a computer vision framework to monitor ecosystem changes related to marine renewable energy (MRE) projects.
The approach automates fish detection and species identification, significantly reducing video analysis time.
This scalable framework enhances ecological monitoring, supporting sustainable development while safeguarding marine ecosystems.

Advancements in Ecosystem Monitoring

The transition to marine renewable energy (MRE) sources like ocean currents and waves is pivotal for low-carbon electricity generation and climate change mitigation. However, introducing MRE devices raises ecological concerns, including potential damage to marine fauna and alterations in habitats. Therefore, it is essential to establish robust environmental baselines and continuously monitor these sites from installation through decommissioning.

Traditional monitoring methods can be invasive and costly, particularly in high-energy aquatic environments. This study explores how advancements in computer vision and artificial intelligence could provide a more effective and scalable solution for monitoring biodiversity and ecosystem dynamics around MRE infrastructure.

Framework Development for Monitoring

Researchers developed a semi-automated framework that utilizes computer vision techniques to analyze underwater video footage collected near a vertical-axis turbine prototype in Puerto Morelos, Mexico. The framework efficiently filters out frames without fish before conducting fish detection and species identification using the YOLOv11 deep learning model, complemented by expert validation against a public species database.

To assess spatial interactions between marine organisms and the MRE device, a monocular depth estimation algorithm was utilized to reconstruct three-dimensional distances from two-dimensional video feed. This method enables researchers to evaluate proximity and collision risks effectively. The framework is designed to be adaptable, allowing its application across various project phases and marine settings.

The case study focused on local fish species, particularly emphasizing the detection of lionfish (Pterois volitans), an invasive species. The framework aims to be user-friendly, utilizing open-access computational tools suitable for environmental monitoring.

Efficiency of Fish Detection and Depth Estimation

This workflow has demonstrated significant efficiency in video analysis, cutting down footage needed for manual review from over 16 hours to slightly more than 1 hour. The YOLO model proved effective in detecting fish presence and species identification, especially for conspicuous or invasive species like lionfish, despite limited training data. However, the ability to identify smaller or less distinct species remains a challenge, necessitating expert oversight and enhanced training datasets.

Additionally, the depth estimation technique has shown effectiveness in measuring distances between fish and the camera, providing valuable data on behavioral responses and the likelihood of collisions. The use of readily deployable video systems combined with computer vision is seen as advantageous for collecting comprehensive ecological data.

The collected data is critical for evaluating species richness, abundance, spatial interactions, and functional diversity before, during, and after the deployment of MRE devices. Notably, the study acknowledges specific limitations such as water conditions affecting video quality and organism behavior.

Implications for Marine Renewable Energy Monitoring

This research establishes that merging advanced image analysis with deep learning presents a viable strategy for monitoring ecological impacts linked to marine energy infrastructure. By automating the detection of marine life and estimating spatial interactions, the approach not only saves time but also sheds light on species behavior and ecosystem modifications.

The scalability and adaptability of the framework facilitate its implementation across different phases of MRE projects, promoting informed mitigation and management strategies. Although challenges exist, the proposed framework provides a foundation for effective, non-invasive environmental monitoring tailored for the complexities of MRE sites.

Refining and expanding the applications of these computer vision techniques remains vital for balancing marine ecosystem integrity with the pursuit of energy production goals.

Journal Reference

Alamillo-Paredes A., Lagunes-Díaz E.G., et al. (2026). A computer vision-based approach to monitor changes in ecosystems associated with marine renewable energy projects. Scientific Reports. DOI: 10.1038/s41598-026-62922-4, http://nature.com/articles/s41598-026-62922-4

The content above is a summary. For more details, see the source article.