Detecção de nematoides na cultura cafeeira a partir de imagens obtidas por aeronave remotamente pilotada

• Autor(a) principal: Vinhal, Geovane Pivêta
• Data de Publicação: 2018
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFU
• Texto Completo: https://repositorio.ufu.br/handle/123456789/23361
• topic: Cafeicultura; Coffee-Growing; Nematoides; Nematodes; ARPs; RPAs; Mapeamento; Mapping; Detecção de Doenças; Detection of Diseases; Estimativa de Teor de Clorofila; Estimation of Chlorophyll Contente; CNPQ::CIENCIAS EXATAS E DA TERRA::GEOCIENCIAS

The coffee crop, for being sustainable, once grown on a large scale, shall be subject to eventual harm as a consequence of the pests incidence and diseases, severely affecting the crop and causing irreparable damages. Among the problems found in coffee plants, nematodes, phytoparasitic that attack the root system of the plant, stands out as being one of the main factors that cause nutritional imbalances, leading to variations in the leaf spectral response. In this context, remote sensing supported by sensors on Remotely Piloted Aircraft (RPAs) is an efficient alternative for the localized management and for combating the dissemination of sick regions, allowing a detailed mapping of the plantation and providing accurate information in timely manner. In view of the above, this work aimed to map a coffee crop infested with nematodes, using an RPA equipped with two sensors, one RGB (red, green and blue) and another multispectral RGN (red, green and infrared channels). For this purpose, a mapping methodology was created by processing techniques of digital image, discriminating the healthy coffee of the infected coffee from samples collected in situ and by the process of not supervised classification by maximum likelihood, which obtained a total accuracy of 81% and kappa coefficient of 0,72 for the RGB image. In a second moment, the concentration levels of chlorophyll content was estimated, considering the radiometry values of the images and correlating them with the chlorophyll samples collected in situ. The RGB image bands showed no correlation, in contrast, in the RGN image, the vegetation index corresponding to the NDVI presented a significant correlation, allowing the creation of empirical models to estimate the chlorophyll A, B and Total with an accuracy of 69,28%, 53,32% and 54,01%, respectively.