Assessing accelerometer thresholds for cow behaviour detection in free stall barns: a statistical analysissis

Abstract

Monitoring daily cow behavioural activities of cows in livestock farms is strategic for improving the herd management. For this reason, IoT techniques and smart sensors are become the most common technological support in barns. The aim of this paper is to validate the use of predefined accelerometer thresholds in timely detecting of cow behavioural activities through the statistical analysis of the data acquired from accelerometers housed in collars. Applying ANOVA and TUKEY tests to the median of the accelerations measured with 4 Hz sampling, the behavioural activities analysed in this study, i.e., feeding, lying, rumination, were found to be discriminable along one or more axes. This could allow the implementation of threshold-based algorithms in the firmware of devices housed in the cow collars.