On Cow Behaviour Classification Methods for Monitoring Grazing Cattle

In recent years, the progressive increase in extensive farming is causing significant environmental impacts. It is estimated that greenhouse gas emissions resulting from grazing livestock constitute approximately 30% of overall anthropogenic emissions. A further damage that should not be underestimated is soil erosion and desertification. The degradation processes associated with the impact of extensive farming on the soil are often due to the adoption of non-rational grazing systems, with consequent compaction phenomena, paths and erosion caused by the activity and behaviour of the animals. Therefore, recognizing the behavioural habits of grazing cows could become a crucial aspect in reducing the impact on the soil. In this article, different methods for classifying the behavioural activities of grazing cows were described. Data were collected by using collars embebbed with accelerometers and two different approaches were compared: one based on the application of neural networks and the other on statistical methods. Points of strength and weakness of the two approaches were discussed and best choice was suggested in relation to the specificities of extensive livestock systems, in terms of site morphology and extension.