Identificação do comportamento bovino com o uso de internet das coisas e inteligência artificial

Abstract

Population growth is not an isolated phenomenon. It is a consequence, and cause, of a number of other factors. As one of the main consequences of the world population increase, we can highlight the increase in food demand, heating and demanding more from several production chains. It is no different with the agricultural productive sector. Large increases in food demand are projected, especially in the consumption of meat and animal derivatives in the coming decades, and this brings new challenges to rural producers. In order to be a support tool in facing these new challenges, the concept of precision livestock has been developed in recent years, where animal monitoring is one of the central points of the concept. Coming to solve welfare and behavior monitoring problems, using sensors, wireless networks and the internet of things, precision livestock farming seeks to provide data on the health and behavior of the herd, mitigating the problem of visual monitoring of the animal. animal. These data can be very useful for the rural producer, as they allow for an anticipation of losses caused by diseases, discomfort and welfare problems that affect the herd, as such a mechanism has the capacity to monitor cattle full time. For this, an electronic device was developed, dressed in the animal, in the form of a collar, which collects information about the behaviors expressed by the animals on the property, through sensors. From the data collected by the device, artificial intelligence algorithms were applied to classify the animal’s behaviors, such as Deep Learning, as well as classic algorithms, such as Multilayer Perceptron, SMO, J48 decision tree and Randon Forest decision tree. In the end, it was observed that the classic algorithms presented good behavior against the Deep Learning approach, highlighting among them the decision trees, which in some cases, showed an accuracy of 93% in the classification of data.