Estimativa e discriminação de áreas de soja [Glycine max L.] no estado do Paraná com dados mono e multitemporais do sensor MODIS

• Ano de defesa: 2014
• Autor(a) principal: Silva Junior, Carlos Antonio da
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual de Maringá; Brasil; Departamento de Agronomia; Programa de Pós-Graduação em Agronomia; UEM; Maringá, PR; Centro de Ciências Agrárias
• Palavras-chave em Português: Soja; Espectrorradiometria; Índices de vegetação; Redes neurais artificiais; Sensoriamento remoto; Geobia; Paraná; Brasil.; Soy; Spectroradiometry; Vegetation indices; Artificial neural networks; Remote sensing; GEOBIA; Brazil.; Ciências Agrárias; Agronomia
• Link de acesso: http://repositorio.uem.br:8080/jspui/handle/1/1235

The monitoring of dynamics vegetation, mainly in agricultural activities with the use of techniques of remote sensing not only has increased research, but also answers a need recognition of the territorial patterns of a local. Satellite images, chiefly from MODIS sensor, show a significant potential in the mapping of extensive agricultural areas in Brazil, once crops as the soybeans are cultivated in the summer with constant presence of clouds. Thus, the images from this sensor allows mapping through different reasons, such as: temporal and spectral resolution, swath and free availability. Therewith, the guiding objective of this research is verify the potential to estimate and map cultivated areas with the soybean crop over MODIS images with and without time-series at the Paraná State. For the characterization of samples from soybean areas, were collected points with GPS distributed in the State. Also were cultivated soybean plants in a greenhouse to correlate laboratory and orbital spectral reading. Time-series EVI and PVI of MODIS images from harvests-year 2010/2011 and 2011/2012 were used for confection of the maps of soybean areas, with routines algorithms of artificial neural networks, oriented analysis in geo-object, main components, overseen classification, partially no overseen and vegetation indexes. Furthermore, was developed the PCEI (Perpendicular Crop Enhancement Index) index based on the soil line and determined by decision tree, to automate the mapping of areas with soybean plants. The maps were evaluated through parameters Kappa and Overall Accuracy with comparison done by Z test (K = 0.05). The results showed that mapping done over the oriented analysis in geo-object, neural networks and PCEI present satisfactory conditions for this purpose. The analysis of time-series used in PCEI index allows distinction from other agricultural crops, been analyzed since the soil prepare until de harvest. Mapping and discrimination of soybean areas at Paraná demonstrated to be viable with MODIS images what that in systematization showed sufficient results of the analyzed parameters.