DETECÇÃO DE RESPOSTAS ESPECTRAIS NA SOJA CAUSADOS POR NEMATOIDES POR MEIO DE SISTEMAS AÉREOS NÃO TRIPULADOS

Abstract

Unmanned aerial systems have become an auxiliary tool in monitoring nematode damage in soybeans. The objectives of this study were to analyze the relationship between the nematodes numbers (P. brachyurus, H. glycines, H. dihystera) and the spectral response of soybean plants; the nematodes radial distribution around the initial focus of infestation; estimate the population of P. brachyurus, H. dihystera and agronomic variables by fusion of multiple sensor data; select spectral bands that best estimate the injury caused by P. brachyurus, H. dihystera and agronomic variables. Initial exploratory flights at the V5 and R1 soybean growth stages were carried out to identify problems, with a Phantom 4, 120 m high RGB sensor and 80% overlaps. The images were orthorectified and two transects were created for each zone from the initial focus and radially positioning additional sampling points at 10, 20, 40 and 80 m from the center. In stages R1 and R5, flights and samplings of soil and root of soybean plants were carried out simultaneously. RGB images were acquired with Phantom 4 sensor. Multispectral images were acquired with Parrot Sequoia and Sentera sensors that capture images in the visible and near infrared spectra. All images were processed on Pix4D and analyzed using Qgis software. The nematodes extraction in the soil and plants root followed the methods of nematodes extraction used in the laboratory. Mixed effects regression models were used to determine the relationship between the nematodes number and spectral reflectance / vegetation indices and the effect of distance from the center of infested plants on the nematodes number. Data fusion was also performed through stepwise multiple linear regression models of the forward model selection between the measured (nematological and agronomic) and estimated variables (19 spectral bands). Soybean spectral responses from low-cost RGB sensors have the potential to predict H. glycines lesions. Pratylenchus brachyurus and H. dihystera in the root of soybean plants were effectively detected and estimated only by fusion of multisensor data. Agronomic variables were estimated with bands that act in the visible and near infrared spectrum. Appropriate sampling protocols are indicated because higher numbers of some nematode species were found outside areas that represent severely stressed plants.