Estimating the Feeding Time of Individual Broilers via Convolutional Neural Network and Image Processing.

Area of Science:

• Animal Science
• Computer Vision
• Agricultural Engineering

Background:

• Feeding behavior is a key indicator of broiler welfare and farm management efficiency.
• Manual monitoring of poultry behavior is labor-intensive and time-consuming, especially in large-scale operations.
• Automated systems are needed to overcome the limitations of visual human observation in poultry farms.

Purpose of the Study:

• To develop and evaluate a convolutional neural network (CNN)-based model for automatically estimating individual broiler feeding times.
• To assess the accuracy of a novel algorithm in recognizing broiler head positions within feeding pans.
• To provide a real-time tool for monitoring broiler feeding behavior in commercial poultry settings.

Main Methods:

• A You Only Look Once (YOLO) model was trained on 1500 labeled images to detect broiler heads.
• A Euclidean distance-based tracking algorithm was implemented to follow detected broiler heads.
• The algorithm determined feeding time by identifying when a broiler's head was positioned inside the feeder.
• Performance was evaluated using three 1-minute labeled videos.

Main Results:

• The developed algorithm achieved an overall feeding time estimation accuracy of 87.3% per broiler visit to the feeding pan.
• The system successfully tracked individual broiler heads in the feeding environment.
• The CNN-based approach demonstrated potential for real-time application.

Conclusions:

• The proposed automated system accurately estimates broiler feeding time, offering a valuable tool for poultry management.
• This technology can enhance the understanding of poultry resource usage and welfare indicators.
• The algorithm provides a scalable and efficient alternative to manual behavioral monitoring in broiler farms.