Moving mean-based algorithm for dairy cow's oestrus detection from uniaxial-accelerometer data acquired in a free-stall barn

In precision livestock farming, accelerometers have been adopted to recognise cow's behavioural activities with the final aim of generating early alarm of illness or specific physiological statuses. Accelerometers are widely used for their low cost and easy integration with other ICT devices and, when they are adopted to acquire data for cow's behavioural activities recognition, models based on acceleration threshold values make it possible to obtain low computational classifiers. In this study, a model based on an acceleration threshold algorithm is proposed in order to detect cow's oestrus activity. In detail, a threshold to predict oestrus onset was determined based on experimental data acquired during specific experimental tests carried out in a free-stall barn for dairy cows. Oestrus onset was validated by using images from a video-recording system. Novelty in the results achieved in this study regards the model features that are suitable for the production of a plug-and-play cow's behaviour monitoring system based on specific firmware-equipped devices.