Identificação de ferrugem na soja por meio de imagens de alta resolução espacial

Abstract

Soybean cultivation has had great economic importance for Brazil, since it is one of the main generators of foreign exchange for the country. However, despite all existing technologies, crops are still constantly being affected by foliar diseases such as Asian soybean rust (Phakopsora pachyrhizi), which is one of the most destructive diseases of soybean, frequently causing significant loss of yield and rapid field-to-field dissemination through urospores dispersed in the air. In order to implement fungicide treatments in a timely manner for effective disease control, it is essential to detect an infection and severity of soybean rust. As a form of primary infection identification, the present research was based on the use of remote sensing recorded by multispectral high spatial resolution image (RapidEye) to detect, discriminate and spatialize a possible occurrence of soybean rust at different levels of severity. For this, a color index and a multispectral characterization were used for pathogen detection in images, and the vegetation indices to be inferred as areas of infestation. From the generated letters, it was possible to characterize spatially as rust areas in intermediate and advanced infection stages, even considering the lack of checkpoints in the field. However, a methodology applied here was not able to detect the early pathogen.