Publications

da Silva, CA; Nanni, MR; Shakir, M; Teodoro, PE; de Oliveira, JF; Cezar, E; de Gois, G; Lima, M; Wojciechowski, JC; Shiratsuchi, LS (2018). Soybean varieties discrimination using non-imaging hyperspectral sensor. INFRARED PHYSICS & TECHNOLOGY, 89, 338-350.

Abstract
Infrared region of electromagnetic spectrum has remarkable applications in crop studies. Infrared along with Red band has been used to develop certain vegetation indices. These indices like NDVI, EVI provide important information on any crop physiological stages. The main objective of this research was to discriminate 4 different soybean varieties (BMX Potencia, NA5909, FT Campo Mourao and Don Mario) using non-imaging hyperspectral sensor. The study was conducted in four agricultural areas in the municipality of Deodapolis (MS), Brazil. For spectral analysis, 2400 field samples were taken from soybean leaves by means of FieldSpec 3 JR spectroradiometer in the range from 350 to 2500 nm. The data were evaluated through multivariate analysis with the whole set of spectral curves isolated by blue, green, red and near infrared wavelengths along with the addition of vegetation indices like (Enhanced Vegetation Index - EVI, Normalized Difference Vegetation Index - NDVI, Green Normalized Difference Vegetation Index - GNDVI, Soil-adjusted Vegetation Index - SAVI, Transformed Vegetation Index - WI and Optimized Soil-Adjusted Vegetation Index - OSAVI). A number of the analysis performed where. discriminant (60 and 80% of the data), simulated discriminant (40 and 20% of data), principal component (PC) and cluster analysis (CA). Discriminant and simulated discriminant analyze presented satisfactory results, with average global hit rates of 99.28 and 98.77%, respectively. The results obtained by PC and CA revealed considerable associations between the evaluated variables and the varieties, which indicated that each variety has a variable that discriminates it more effectively in relation to the others. There was great variation in the sample size (number of leaves) for estimating the mean of variables. However, it was possible to observe that 200 leaves allow to obtain a maximum error of 2% in relation to the mean. (C) 2018 Elsevier B.V. All rights reserved.

DOI:
10.1016/j.infrared.2018.01.027

ISSN:
1350-4495