Avaliação de características agronômicas em soja por sensor ativo de vegetação e câmera multiespectral embarcada em aeronave remotamente pilotada

• Autor(a) principal: Vareiro, Raphael Borgias
• Data de Publicação: 2020
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Manancial - Repositório Digital da UFSM
• dARK ID: ark:/26339/001300000wgz1
• Texto Completo: http://repositorio.ufsm.br/handle/1/23561
• topic: Agricultura de precisão; Sensoriamento remoto; Índice de vegetação; Sequoia; Phantom 4; GreenSeeker; Precision agriculture; Remote sensing; Vegetation index; CNPQ::CIENCIAS AGRARIAS::AGRONOMIA

Soy is one of the main national agricultural commodities, and, according to projections, Brazil will be, in 2020, the largest producer of this grain. As it is inserted in a highly competitive market, agricultural producers are looking for alternatives to obtain better productive and economic performance. Currently adopting precision agriculture (PA) is one of the options to obtain the best results in agricultural production, working with practices and technologies that allow cost reduction, avoiding waste and acting with precision in problems. Remote sensing (RS) is one of the PA techniques in which it obtains the reflectance of an object without having physical contact with it, making it possible to calculate the vegetation indices (VI), which highlight the specific vegetation behavior. There are several sensors available today, among them, multispectral cameras and active vegetation sensors. Multispectral cameras can be loaded on remotely piloted aircraft (RPA), which make it possible to obtain a high degree of detail, due to the high spatial and temporal resolution. Infrared (IR) sensors, on the other hand, provide the achievement of an IR acting at field level. Thus, the objective was, with the use of a Sequoia multispectral camera embedded in an RPA and also with the use of the GreenSeeker vegetation sensor, to assess biomass, plant height and productivity in soybean culture. The experiment was carried out in the 2018/2019 harvest, in the experimental area of the Polytechnic College of Federal University of Santa Maria. To provide variability, three plant populations (12, 24 and 36 plants/m²) were used, in a randomized block design, with four replications. The surveys with Sequoia and GreenSeeker were carried out in stages V10, R3 and R5.3. In the same stages, biomass (kiln-dried) was collected, the height of the plants was measured and in R9, manual harvesting to measure productivity. Statistical analyzes were performed by Pearson's correlation between the VI (GNDVI, MPRI, NDRE and NDVI) and the variables biomass, plant height and productivity. The V10 was the stage in which the best results were observed, in which all VI of both sensors achieved significant correlations (strong and very strong) with biomass and height. From R3 onwards, saturation of the VI was observed among plant populations/m². Even so, NDRE was the only VI to obtain a significant correlation with shoot biomass. in one of the reproductive stages (R3). In R5.3, there was no correlation between vegetation indexes and any agronomic variables evaluated. There was no significant correlation of productivity with any of the VI, in any phenological stage. The MPRI achieved a performance similar to the other multispectral VI in the V10 stage, which makes it possible to obtain a potential VI using a visible RGB camera. It is understood that the plasticity of the soybean plant and the greater canopy closure from R3 onwards made it difficult to obtain better results in the correlations.