Aplicação de algoritmos de aprendizagem supervisionada de máquina em sistemas embarcados no auxílio à aplicação de defensivos agrícolas

Abstract

Machine learning has been increasingly used to improve weather monitoring and forecasting. It uses a large volume of data to establish patterns and predictively categorize new input data. Thus, data collected by automatic agro­meteorological stations have a significant impact on the collection and processing of data, as they present specific conditions for the execution of certain tasks inherent to the climate, such as, for example, the application of agricultural defensives, where the drift phenomenon. The present work aimed to address the possibility of applying machine learning and IoT concepts in local meteorological stations, aiming to minimize the impact of drift on agricultural production. Discussions are presented on the operation and development of machine learning systems in its embedded system, aiming to improve the accuracy and collection of information and using them as an aid to the application of agricultural pesticides. In this way, a station was developed with an embedded system for the collection of climatic data in real time, using machine learning models to identify the ideal meteorological conditions for the application of pesticides, which works as a WI­FI signal repeater. . A monitoring platform was also developed for the collected meteorological data, and the presentation of the respective answers of the machine learning algorithms as an aid to the spraying, called i­Cida. Finally, the success rates of the algorithms in laboratory tests, in the rural partner company and the results of the algorithms in the microcontroller in the panel of the developed application are presented, in addition to comparing the achieved levels of IoT by the project.