Development of a New Wearable 3D Sensor Node and Innovative Open Classification System for Dairy Cows' Behavior.

Area of Science:

• Animal Science
• Agricultural Engineering
• Biotechnology

Background:

• Dairy cattle monitoring is crucial for detecting health and welfare issues, enabling early intervention.
• Existing commercial sensors often lack the raw, open data necessary for detailed behavioral research.
• There is a need for customizable, open-access systems for precise cattle behavior classification.

Purpose of the Study:

• To develop an open and customizable sensor system for classifying dairy cattle behaviors.
• To provide researchers with sufficient raw and open data for detailed analysis.
• To enhance the potential for data processing through integration with other information sources.

Main Methods:

• A 3D accelerometer device and host-board were integrated into a collar for dairy cows.
• The system operated in acquisition mode (mobile app for data collection) and operating mode (data processing and cloud transmission).
• The Decision Tree algorithm was selected after evaluating multiple models using 108 hours of data from 32 cows, achieving 85.12% accuracy in classifying standing, lying, eating, and ruminating behaviors.

Main Results:

• The developed system successfully classified four key dairy cattle behaviors: standing, lying, eating, and ruminating.
• The Decision Tree algorithm achieved an average accuracy of 85.12% in behavior classification.
• The system's open nature allows for future functional additions and integration with other sensors.

Conclusions:

• The developed open-source system provides a valuable tool for monitoring dairy cattle behavior and improving animal welfare.
• The system's flexibility enables customization and integration, enhancing its utility in research and farm management.
• Accurate classification of behaviors like standing, lying, eating, and ruminating supports early detection of health and welfare concerns in dairy herds.