Classificação de genótipos de soja quanto ao desempenho fisiológico utilizando variáveis espectrais

• Ano de defesa: 2022
• Autor(a) principal: REGIMAR GARCIA DOS SANTOS
• Orientador(a): Larissa Pereira Ribeiro Teodoro
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Fundação Universidade Federal de Mato Grosso do Sul
• País: Brasil
• Palavras-chave em Português: Predição, Caracteres fisiológicos, Agronômicos, Soja, Índices de vegetação.
• Link de acesso: https://repositorio.ufms.br/handle/123456789/4334

The physiological characters of the soybean crop are correlated with its productivity, which has been proven through studies of genetic progress. The measurement of these characters requires complex equipment, specialized labor, in addition to being costly and time consuming. Problems that can be solved using spectral bands and vegetation indices associated with remote sensing. The objective of this work was to identify the best machine learning technique for classifying F2 soybean population clusters based on their physiological characteristics using spectral bands and vegetation indices as input variables. The experiment was carried out in the 2019/2020 agricultural year, in the municipality of Chapadão do Sul, MS. Using a drone, the values of spectral bands (SB) and vegetation indices (IV) were collected from 194 soybean populations in F2. The physiological characteristics evaluated were: net photosynthesis(A),stomatal conductance (gs), internal concentration of CO2 (Ci), transpiration (E) and water use efficiency (USA). Using the k-means technique the samples were divided into two clusters, using Principal Components (PCA), the samples were partitioned into two groups based on their physiological behavior. SB+IV, SB only and IV only were evaluated as input variables. Using SB+IV, the technique with the highest classification capacity was the artificial neural networks (ANN) with 66.34% of correct classifications (CC), the J48 algorithm presented the best result using only SB (69.87% CC) and Logistic Regression (RL) obtained better response when using only IV's as input variable (68.95% CC). The results obtained demonstrate that the best way to make the classification is using only the SB as input variables in the J48 algorithm, reducing the time required and the chances of error during the step of calculating the vegetation indices.