Analysis of Accelerometer and GPS Data for Cattle Behaviour Identification and Anomalous Events Detection.

Area of Science:

• Agricultural Engineering
• Animal Science
• Machine Learning

Background:

• Cattle behavior monitoring is crucial for farm management and welfare.
• Accurate classification of behaviors like grazing and ruminating aids in assessing pasture use.

Purpose of the Study:

• To develop and validate a method for classifying cattle behavioral patterns using sensor data.
• To assess the accuracy of machine learning models in distinguishing between grazing, ruminating, laying, and standing behaviors.

Main Methods:

• Equipping cattle with 3-D accelerometers and GPS sensors.
• Extracting time and frequency domain features from accelerometer data.
• Training a random forest classifier with video-matched activity patterns.
• Utilizing k-medoids clustering for GPS-based herd location tracking.

Main Results:

• The accelerometer-based classification achieved high accuracy, with the best performance (0.93) for grazing detection.
• The combined sensor approach effectively tracks herd location and spatial distribution.
• The method shows potential for monitoring sustainable pasture consumption and detecting anomalies.

Conclusions:

• The proposed sensor-based method accurately classifies cattle behaviors, particularly grazing.
• This technology offers a valuable tool for farm management, enabling monitoring of pasture use and early detection of issues.
• Further validation across diverse farm environments is recommended to consolidate the strategy.