Classification of dairy cows’ behavior by energy-efficient sensor

Precision Livestock farming (PLF) is a combination of the use of biosensors and a data mining processes to ensure automatic and individual dairy cow management. It essentially allows a better monitoring of health status, morphology, reproduction, and welfare of dairy cows. Collecting behavioral activities of dairy cows has the potential to improve the productivity and the longevity. Behavioral monitoring with attached sensors helps to detect several diseases at an early stage before clinical signs appear and thus facilitates treatment. This paper proposes a new technique based on the time-driven embedded sensor to continuously measure and classify the behavior of dairy cows housed in free stall. This sensor is based on an Inertial Measurement Unit and is attached to the back of dairy cow. A machine learning model has been elaborated and implemented in the sensor. Univariate and multivariate models were used to define threshold values for the decision trees used to classify the behavior of dairy cows. The obtained results showed that our results in terms of rates classification and energy consumption are very promising compared to other research works.