Estimativa da produtividade do milho a partir de imagens de alta resolução espacial

Abstract

Recently, remote sensing is becoming more important due to the cost/benefit ratio and it's used even more to monitor the tillage and plantations. One of the technologies that has been rigorously promoting the remoting sensing in precision agriculture are the orbitals images of high resolution, that can monitor agricultural plantations and assist in decision-making processes in the main agronomic environments. The speed of this technology has made it one of the most important technologies in precision agriculture nowadays. One of the goals of precision agriculture is to help to determine agronomic parameters that promote the field activities and raise the productivity of the farming. To estimate the productivity of the farming, it is necessary to understand the conduct of each analyzed culture to find the characteristics of possible patterns of each plantation. Therefore, the main challenge comes with cultures with more specific conduct, like the white culture. For this, vegetation indices combined with spectral bands to estimate productivity of white culture in a determined area, so using machine learning algorithms, it's intended to present the application of the estimation models through spatial distribution maps of productivity. The studies conclude that the best vegetative stadia to estimate the corn productivity is the R3, through the Multilayer Perceptron that showed an accuracy about 78% and a correlation index of 0,57.