Remote sensing and computer vision for marine aquaculture.

Area of Science:

• Aquaculture
• Remote Sensing
• Computer Vision

Background:

• Aquaculture production data are sparse, hindering trend analysis and risk assessment.
• Existing data are self-reported and aggregated, limiting detailed understanding.
• Effective monitoring is crucial for managing this rapidly growing industry.

Purpose of the Study:

• To develop a computer vision model for identifying marine aquaculture cages from imagery.
• To generate a spatially explicit dataset of finfish production locations.
• To provide a scalable and adaptable method for monitoring aquaculture.

Main Methods:

• Trained a computer vision model using manual surveys of remote sensing imagery.
• Identified marine aquaculture cages from aerial and satellite imagery.
• Generated a dataset of finfish production locations (2000-2021) in the French Mediterranean.

Main Results:

• Created a dataset of 4010 marine finfish cages.
• Demonstrated an adaptable, cost-effective approach for aquaculture surveys.
• Enabled independent production estimates and uncertainty quantification.

Conclusions:

• The developed method is efficient, scalable, and adaptable for monitoring aquaculture.
• Improves the speed and reliability of aquaculture surveys.
• Supports downstream analyses for researchers and regulators.