Robust Tracking and Quantification of Open-Beak Behavior for Heat Stress Assessment in Multi-Tier Caged Rearing...

Area of Science:

• Animal Science
• Computer Vision
• Biotechnology

Background:

• Panting is a key indicator of heat stress in chickens, but current vision-based methods struggle to capture its continuous nature.
• Assessing panting based on instantaneous beak opening in single frames is unreliable for understanding heat stress levels.

Purpose of the Study:

• To develop a robust method for quantifying chicken heat stress by analyzing sustained panting behavior.
• To introduce the Cumulative Panting Ratio (CPR) as a reliable metric for continuous heat stress assessment.

Main Methods:

• A cascaded Detection-Tracking-Classification pipeline was developed to decouple body tracking from beak behavior analysis.
• A Near-Online Trajectory Stitcher was implemented to ensure continuous tracking of chickens despite platform movement.
• The Cumulative Panting Ratio (CPR) was proposed, accumulating open-beak events within a sliding window for sustained behavior quantification.

Main Results:

• The proposed method demonstrated robust tracking continuity and stable performance in quantifying sustained open-beak behavior.
• Compared to existing methods, the new approach reduced identity switches and improved IDF1 score by 2.43%.
• System throughput was significantly increased to 51.43 FPS with TensorRT acceleration and parallel processing, without sacrificing accuracy.

Conclusions:

• The Cumulative Panting Ratio (CPR) offers a practical and accurate solution for quantitative heat stress assessment in chickens.
• This work supports the development of advanced physiological state monitoring systems for precision livestock farming.
• The developed vision-based system enhances the reliability and efficiency of monitoring animal welfare and productivity.