Analysis of the Drinking Behavior of Beef Cattle Using Computer Vision

Area of Science:

• Animal Science
• Computer Vision
• Machine Learning

Background:

• Monitoring animal drinking behavior is crucial for livestock health and well-being.
• Traditional methods for measuring drinking time are labor-intensive and challenging for large-scale livestock production.
• Computer vision offers a potential solution for automated and efficient monitoring.

Purpose of the Study:

• To develop a computer vision system for monitoring beef cattle drinking behavior.
• To utilize low-cost camera systems for data acquisition.
• To enable automated analysis of animal welfare through behavioral monitoring.

Main Methods:

• Developed a data acquisition system with an RGB camera and ultrasonic sensor.
• Employed DeepLabCut, a deep learning architecture, for tracking key cattle body parts (head-ear-neck).
• Utilized a long short-term memory (LSTM) model to classify drinking and non-drinking periods from extracted key points.

Main Results:

• The developed system accurately tracked beef cattle key body parts.
• The LSTM model achieved 97.35% accuracy in classifying drinking and non-drinking periods during testing.
• A total of 70 videos were used for training and testing, with 8 for validation.

Conclusions:

• The computer vision system effectively monitors beef cattle drinking behavior.
• This technology can enhance farmers' capabilities in monitoring animal health and well-being.
• The findings address immediate needs in livestock farming for efficient behavioral monitoring.