Related Experiment Video
Updated: Dec 1, 2025
10:07
Thermal Imaging to Study Stress Non-invasively in Unrestrained Birds
Published on: November 6, 2015
13.8K
Non-Invasive Sheep Biometrics Obtained by Computer Vision Algorithms and Machine Learning Modeling Using Integrated
Sigfredo Fuentes1, Claudia Gonzalez Viejo1, Surinder S Chauhan2
1Digital Agriculture, Food and Wine Sciences Group, School of Agriculture and Food, Faculty of Veterinary and Agricultural Sciences, The University of Melbourne, Parkville, VIC 3010, Australia.
Sensors (Basel, Switzerland)
|November 11, 2020
Summary
This study introduces an AI-powered, non-contact system to monitor sheep welfare by assessing heat stress. The system accurately estimates respiration rate and heart rate using thermal and RGB videos, improving animal welfare assessments.
Area of Science:
- Animal Science
- Artificial Intelligence
- Biotechnology
Background:
- Live sheep export raises public concern regarding animal welfare.
- Assessing animal welfare, particularly heat stress, during transport and on farms is crucial.
- Non-contact methods are needed for efficient and objective animal monitoring.
Purpose of the Study:
- To develop and validate a non-contact biometric system using artificial intelligence (AI) for assessing heat stress in sheep.
- To automate animal welfare assessment in farm and transport settings.
- To extract physiological parameters like heart rate (HR) and respiration rate (RR) from video data.
Main Methods:
- Infrared thermal videos (IRTV) were used to extract skin temperature from sheep's head features via automated tracking.
- RGB videos were processed to estimate HR (beats per minute) and RR (breaths per minute) using luminosity and CIELAB color space analysis.
- Supervised machine learning (ML) models were developed: a classification model for RR frequency levels and a regression model for HR and RR estimation.
Main Results:
- The ML classification model (Model 1) achieved 96% overall accuracy in estimating respiration rate frequency levels.
- The ML regression model (Model 2) accurately estimated HR and RR, showing a high correlation (R = 0.94) and a slope of 0.76.
- Both models demonstrated high accuracy without statistical signs of overfitting, indicating robust performance.
Conclusions:
- The developed non-contact AI system effectively assesses sheep heat stress using thermal and RGB imaging.
- The system provides accurate, automated estimations of respiration and heart rates, crucial for welfare monitoring.
- This technology holds potential for improving animal welfare management in live sheep export and farming.
