Uso de aeronaves remotamente pilotadas na detecção do bicho mineiro nos cafeeiros

Abstract

Remote sensing (SR) is a technique whose fundamental principle is to obtain data from an object without necessarily having direct contact. short time and cover large areas compared to conventional field methods. Therefore, detecting pests such as the coffee leaf miner (Leucoptera coffeella) becomes a great challenge, since it is a phytosanitary problem of coffee plants and is important for Brazilian agribusiness. In this context, the objective is monitoring for its detection, using a remotely piloted aircraft (RPAs) with RGB cameras and obtaining high resolution images combined with machine learning algorithms. The experiment was carried out on the farm called Barro preto, located in the city of Lavras, state of Minas Gerais, Brazil. The results achieved were the generation of 4 models derived from machine learning, being the ‘ranger’ model derived from Random Forest with 90% accuracy in detecting the presence of leaf miner infestation through RGB image, with sensitivity adjustment 88% of detecting the object in the field and specificity of 91% demonstrating that there is a real possibility of managing and managing the coffee crop, through low-cost RGB images, in relation to the miner infestation.