天美传媒

Abstract

The use of automated measurement to monitor the behaviour of animals is now widespread. Key features that have been identified include restlessness (as an indication of oestrus), eating and rumination. The paper presents the development and evaluation of eating and rumination signatures derived by processing signals from a three-axis accelerometer, the foundation for a range of decision support services optimising farm operations through alerting of the early onset of illnesses e.g. the basis for a cloud based nutrition service. The methodology for determining eating and rumination signatures using features within accelerometer raw measurement data is detailed. Furthermore evidence is provided that the signatures are a highly accurate and robust determination of these cow states. Using a Hidden Markov state transition model that determines the likelihood of a transition between behavioural states e.g. when the animal transitions between eating and ruminating, an overall prediction accuracy of better than 90% is obtained validated through commercial trials. The engineering approach adopted is designed to be compatible with a low power processing platform while lends itself to extended lifetime operation (7 years) using two AA batteries. Case studies on commercial farms are presented where the Silent Herdsman platform has flagged significant changes in eating and rumination patterns, a clear indication of the onset a detrimental animal welfare condition before any clinical signs become visible.

Biography