Identificação de Colletotrichum gossypii e Colletotrichum gossypii var. cephalosporioides em sementes de algodoeiro usando imagens hiperespectrais no infravermelho próximo.

• Ano de defesa: 2019
• Autor(a) principal: Jesus, Hanna Ibiapina de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Fitotecnia e Ciências Ambientais; Programa de Pós-Graduação em Agronomia; UFPB
• Palavras-chave em Português: Antracnose.; HSI-NIR.; Ramulose.; CNPQ::CIENCIAS AGRARIAS::AGRONOMIA
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/17404

Brazil is consolidating itself in the international market as one of the world’s leading cotton fiber producers and exporters. Despite the good performance, phytosanitary problems, in particular, are a major obstacle, in which one of the main sources of pathogens dispersions are contaminated seeds. Methodological problems in detection and differentiation of the fungal species Colletotrichum gossypii (CG) and Colletotrichum gossypii var. cephalosporioides (CGC) in cotton seeds have been subject for researches, because the species morphological structures similarity causes ambiguous results that contributes for commercialization of infected seeds. In this context, this study was aimed to develop a methodology for the classification of CG and CGC in cotton seeds, using hyperspectral imaging near infrared (HIS- NIR) in association with data multivariate analysis. Therefore, cotton seeds of BRS 286 cultivar were contaminated with 5 CG isolates and 15 CGC isolates, and were submitted to seed health testing, using the Blotter test method. After the incubation period, hyperspectral images of the seeds were taken. The spectra in the range between 1000 to 2500 nm were pre-processed with Savitzky-Golay first-order derivative. An exploratory data analysis was executed using a PCA, posteriorly, samples classification was made by the development of a PLS-DA model, which correctly predicted 86.5% of CG class and 81.6% of CGC class. In external samples’ prediction, the correct prediction percentage was variable between samples and, possibly, it is related to species variability. The PLS-DA model performance indicates that this method allows CG and CGC identification, however, samples with high rate of misclassification should be included in calibration set, and model construction adjustments are necessary for improvement of the fungal species classification in cotton seeds.