| Autor: |
Guilherme Teixeira Crusiol, Luís, Braga, Patricia, Rafael Nanni, Marcos, Furlanetto, Renato Herrig, Sibaldelli, Rubson Natal Ribeiro, Cezar, Everson, Sun, Liang, Foloni, José Salvador Simonetto, Mertz-Henning, Liliane Marcia, Lima Nepomuceno, Alexandre, Neumaier, Norman, Bouças Farias, José Renato |
| Předmět: |
|
| Zdroj: |
International Journal of Remote Sensing; Nov 2021, Vol. 42 Issue 21, p8165-8184, 20p |
| Abstrakt: |
Because of the major international issue of intellectual property rights over seed technology, time-efficient methodologies that could characterize and monitor soybean genotypes are increasingly needed as they contribute to royalty collection and better management of seed genetics, leading to more efficient production strategies. This research aims at developing models capable of spectrally classifying soybean genotypes, at leaf scale, under different levels of water availability. Leaf reflectance spectra were collected using the Fieldspec 3 Jr. spectroradiometer, a hyperspectral proximal sensor. Two experiments were conducted at Embrapa Soja (Brazilian Agricultural Research Corporation). Experiment I was carried out in a plant growth chamber, using 12 soybean genotypes with different tolerance levels to drought, submitted to irrigation and water deficit treatments, where 214 spectral samples were collected. Experiment II was conducted in a greenhouse, using 164 soybean genotypes (from 15 different countries) with different responses to drought, submitted to irrigation and water deficit treatments, where 6,046 spectral samples were collected. Principal component analysis (PCA) was used to explore the qualitative spectral differences among the evaluated genotypes within each spectral dataset, and the stepwise procedure was carried out to select the wavelengths that best discriminate the soybean genotypes. Linear discriminant analysis (LDA) was used to obtain a model of classification of each leaf-based reflectance spectrum for each soybean genotype by its spectral behaviour. The acquired leaf reflectance was similar among soybean genotypes. The PCA explained over 96% of the spectral variance in the first three principal components, and the stepwise procedure selected 70 and 410 spectral bands in each dataset. In Experiment I, LDA presented an accuracy of 100%. In Experiment II, the LDA classified 149 genotypes with an accuracy of over 90%. The obtained results demonstrate the great potential of the spectral classification of soybean genotypes at leaf scale, regardless of the water status of the plants. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
