Using Thermal Signature to Evaluate Heat Stress Levels in Laying Hens with a Machine-Learning-Based Classifier.

Area of Science:

• Animal Science
• Agricultural Engineering
• Biotechnology

Background:

• Animal welfare and performance are influenced by environmental factors like heat stress.
• Infrared thermography offers a non-invasive method to monitor animal physiology.
• Developing accurate methods to detect heat stress is crucial for poultry production.

Purpose of the Study:

• To propose and validate the thermal signature method for extracting features from infrared thermography data.
• To construct computational models for classifying heat stress levels in laying hens.
• To evaluate the performance of different machine learning models for this classification task.

Main Methods:

• Infrared thermography was used to capture surface temperature data from laying hens under heat stress and thermal comfort conditions.
• Thermal signatures were extracted from specific body regions (face, eye, wattle, comb, leg, foot).
• Machine learning classifiers (Random Forest, Random Tree, Multilayer Perceptron, K-Nearest Neighbors, Logistic Regression) were trained using thermal signatures and rectal temperature labels.

Main Results:

• No significant differences in thermal signatures were observed between the two laying hen strains.
• Both rectal temperature and thermal signatures effectively indicated heat stress and thermal comfort.
• The Random Forest model using the face area achieved the highest classification performance (89.0%), followed by the wattle area (88.3%).

Conclusions:

• The thermal signature method is a valid approach for feature extraction from infrared thermography.
• Combining infrared thermography with machine learning provides a promising tool for monitoring thermal stress in laying hens.
• Specific body areas, like the face and wattle, are significant indicators of heat stress in poultry.