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

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