Periodic Reporting for period 2 - Bluegrove Welfare Shield (Sustainable Seafood Production)
Periodo di rendicontazione: 2023-07-01 al 2024-12-31
Bluegrove’s WelfareShield is a project by CageEye, co-funded by the European Innovation Council, aimed at addressing a critical, unresolved challenge in aquaculture: fish welfare.
This initiative is developing the most advanced salmon welfare monitoring and prediction system available. At its core, the system relies on echosounders to continuously monitor fish behavior 24/7. Additionally, feeding cameras provide complementary visual data, further enhancing the accuracy of welfare assessments. The collected real-time data is combined with environmental models (e.g. water temperature, oxygen levels, sunlight exposure, and weather conditions) and farming parameters to identify patterns associated with welfare concerns.
The gathered insights are displayed on an intuitive dashboard, accessible both locally and remotely, providing live analysis of fish behavior, appetite tracking, and welfare indicators for potential risks. By enabling early detection of welfare issues, WelfareShield empowers farmers to take preventive measures, helping to reduce fish mortality and improve growth. Additionally, it helps salmon producers implement best practices and benchmark their performance against the most productive fish cages
This initiative is developing the most advanced salmon welfare monitoring and prediction system available. At its core, the system relies on echosounders to continuously monitor fish behavior 24/7. Additionally, feeding cameras provide complementary visual data, further enhancing the accuracy of welfare assessments. The collected real-time data is combined with environmental models (e.g. water temperature, oxygen levels, sunlight exposure, and weather conditions) and farming parameters to identify patterns associated with welfare concerns.
The gathered insights are displayed on an intuitive dashboard, accessible both locally and remotely, providing live analysis of fish behavior, appetite tracking, and welfare indicators for potential risks. By enabling early detection of welfare issues, WelfareShield empowers farmers to take preventive measures, helping to reduce fish mortality and improve growth. Additionally, it helps salmon producers implement best practices and benchmark their performance against the most productive fish cages
The Bluegrove Welfare Shield project has successfully advanced fish welfare monitoring and optimization through innovative AI and hyd roacoustic technology. The main achievements are:
1. Development of a Real-Time Appetite Monitoring System
○ Implemented deep learning-based pellet detection using YOLOv8.
○ Implemented deep learning-based feeding response model for hydroacoustic data.
○ Integrated pellet detection into acoustic feeding response model for appetite estimation.
○ Analyzed 25,000 feeding days to develop an appetite-based welfare metric.
2. AI-Driven Fish Welfare and Stress Monitoring
○ Created a machine learning health model based on appetite patterns.
○ Developed heuristic models for fish movement analysis to detect stress and abnormal behavior.
○ Established a statistical welfare scoring system combining multiple indicators.
3. Industrialization and Deployment of the System
○ Designed a fully digital hydroacoustic sensor with improved signal processing.
○ Installed and validated the system in 13 sites and 60 production cages across Norway.
○ Shifted focus to integrating standard feeding cameras for enhanced scalability.
The project successfully delivered an AI-enhanced fish welfare monitoring system that integrates hydroacoustic and visual data to optimize feeding strategies and improve fish health. The technology has been validated in commercial production environments, demonstrating significant potential to enhance fish welfare, reduce stress-related losses, and improve aquaculture efficiency. The shift towards third-party camera integration ensures greater scalability and market readiness, making the Bluegrove Welfare Shield a promising tool for the future of sustainable fish farming
1. Development of a Real-Time Appetite Monitoring System
○ Implemented deep learning-based pellet detection using YOLOv8.
○ Implemented deep learning-based feeding response model for hydroacoustic data.
○ Integrated pellet detection into acoustic feeding response model for appetite estimation.
○ Analyzed 25,000 feeding days to develop an appetite-based welfare metric.
2. AI-Driven Fish Welfare and Stress Monitoring
○ Created a machine learning health model based on appetite patterns.
○ Developed heuristic models for fish movement analysis to detect stress and abnormal behavior.
○ Established a statistical welfare scoring system combining multiple indicators.
3. Industrialization and Deployment of the System
○ Designed a fully digital hydroacoustic sensor with improved signal processing.
○ Installed and validated the system in 13 sites and 60 production cages across Norway.
○ Shifted focus to integrating standard feeding cameras for enhanced scalability.
The project successfully delivered an AI-enhanced fish welfare monitoring system that integrates hydroacoustic and visual data to optimize feeding strategies and improve fish health. The technology has been validated in commercial production environments, demonstrating significant potential to enhance fish welfare, reduce stress-related losses, and improve aquaculture efficiency. The shift towards third-party camera integration ensures greater scalability and market readiness, making the Bluegrove Welfare Shield a promising tool for the future of sustainable fish farming
The Bluegrove Welfare Shield project has set a new benchmark in fish welfare management, integrating AI, hydroacoustics, and real-time monitoring to optimize feeding and welfare strategies. This technology has been successfully validated in commercial production environments, paving the way for scalable and impactful applications in modern aquaculture through the following innovations beyond the state of the art.
Advanced Appetite Estimation and AI-Powered Welfare Monitoring
● Multi-modal appetite estimation using deep learning for hydroacoustic feeding response and pellet detection in feeding cameras to track uneaten feed in real time and optimize feeding strategies.
● Cloud-based analytics platform for structured meal data, supporting dynamic appetite scoring and real-time feeding optimization.
● Hydroacoustic behavioral classifiers identifying four key behavioral signatures related to welfare; stress response, activity, diurnal rhythm and abnormal migration patterns.
● Integrated statistical welfare scoring and visualization system, combining behavioral, environmental, and mortality data with a traffic light-based welfare assessment for intuitive fish health monitoring.
● FISHWELL and LAKSVEL-based metrics, providing an objective and standardized approach to fish health assessment.
Hardware Optimization and Sensor Enhancements
● Fully digital hydroacoustic sensor with high signal-to-noise ratio for precise fish biomass monitoring.
● Next-generation cage system, integrating hydroacoustic processing with camera data.
Advanced Appetite Estimation and AI-Powered Welfare Monitoring
● Multi-modal appetite estimation using deep learning for hydroacoustic feeding response and pellet detection in feeding cameras to track uneaten feed in real time and optimize feeding strategies.
● Cloud-based analytics platform for structured meal data, supporting dynamic appetite scoring and real-time feeding optimization.
● Hydroacoustic behavioral classifiers identifying four key behavioral signatures related to welfare; stress response, activity, diurnal rhythm and abnormal migration patterns.
● Integrated statistical welfare scoring and visualization system, combining behavioral, environmental, and mortality data with a traffic light-based welfare assessment for intuitive fish health monitoring.
● FISHWELL and LAKSVEL-based metrics, providing an objective and standardized approach to fish health assessment.
Hardware Optimization and Sensor Enhancements
● Fully digital hydroacoustic sensor with high signal-to-noise ratio for precise fish biomass monitoring.
● Next-generation cage system, integrating hydroacoustic processing with camera data.