Development of an automatic classification system for eating, ruminating and resting behavior of cattle using an accelerometer

Abstract

AbstractMonitoring behavior of grazing animals is important for the management of grazing systems. We developed a new automatic classification system for eating, ruminating and resting activities of cattle using a three‐axis microelectromechanical systems (MEMS) accelerometer. We fitted the accelerometer to a Holstein cow in a tie‐stall barn and to a Japanese Black cow at pasture, and measured their under‐jaw accelerations at 1‐s intervals. The behavior of the animals was also video‐recorded. The raw acceleration data was processed to create 12 variables: the mean, variance and inverse coefficient of variation (ICV; mean/standard deviation) per minute for the x‐, y‐ and z‐axis and the resultant. Quadratic discriminant analysis (QDA) was employed to classify eating, ruminating and resting activities, using 11 combinations of the variables as explanatory variables. In all axes and their resultant, approximately 99.6% of the raw acceleration values ranged between –19.6 m s−2 (–2 G) and 19.6 m s−2 (2 G), with an amplitude tendency of eating > ruminating > resting. Seven combinations of the variables produced total percent correct discriminations exceeding 90% in both tie‐stalled and grazing cows. Overall, the highest discriminant score was obtained in the combination of the ‘Means and ICV’. Our results demonstrate that processing acceleration data with QDA is effective in statistically classifying eating, ruminating and resting activities of cattle.