Espectroscopia no infravermelho próximo para discriminação de espécies de cordas-de-viola
Carregando...
Data
2021-04-15
Autores
Braga, Andreísa Flores [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
A ocorrência de plantas daninhas em áreas agrícolas é um dos principais fatores que limitam a produtividade e aumentam os custos de produção. O manejo específico destas espécies tem ganhado destaque nos últimos anos, principalmente pelo avanço das tecnologias e por apresentar uma alternativa à redução do uso de herbicidas e exposição ambiental. O grande desafio para o manejo específico de plantas daninhas é identificar as espécies alvo em meio às plantas cultivadas. Neste contexto, diversas tecnologias têm sido estudadas para identificação e reconhecimento destas espécies em campo. A espectroscopia no infravermelho próximo (NIRS) é bastante empregada em análises de qualidade e caracterização de alimentos e solos, mas apesar de ser uma técnica bastante difundida, ainda são poucos os estudos relacionados ao seu uso para identificação/discriminação de plantas daninhas. O objetivo deste trabalho foi verificar o potencial discriminatório da espectroscopia NIR associado a análises multivariadas em espécies de cordas-de-viola e estádios de crescimento. Para isso, realizou-se experimentos com os seguintes objetivos: (I) discriminar as espécies de Ipomoea hederifolia, Ipomoea nil e Merremia aegyptia, (II) discriminar as espécies de I. hederifolia e M. aegyptia, considerando três estádios de crescimento. A semeadura das espécies foi realizada em vasos plásticos com capacidade para 0.5 litros, preenchidos com areia de rio lavada e peneirada. As espécies foram cultivadas em ambiente aberto e irrigadas diariamente com água e solução nutritiva. No dia da coleta dos espectros, as plantas foram caracterizadas quanto ao teor de água e ao estádio de crescimento. Os espectros NIRS foram adquiridos em absorbância na faixa espectral de 10.000 a 4.000 cm-1 (1.000 a 2.500 nm). No momento da coleta dos espectros NIR, a temperatura do ambiente foi mantida a 23 °C ± 1 e a leitura dos espectros foi realizada na parte adaxial das folhas frescas, posicionadas sobre um fundo preto. Os dados espectrais foram analisados usando o programa Unscrambler ®. Pré-processamentos e faixas espectrais foram testados e os dados espectrais foram submetidos as análises de componente principal (PCA); discriminante linear de componentes principais (PC-LDA) e apenas no primeiro experimento, a análise discriminante por mínimos quadrados parciais (PLS-DA). O estudo das faixas espectrais bem como os pré-processamentos foram relevantes e promoveram melhorias na capacidade de acerto dos modelos. Os valores de acurácia no primeiro experimento foram de até 99,26, 98,52 e 98,73%, para I. hederifolia, I. nil e M. aegyptia, respectivamente. No segundo experimento as melhores faixas para discriminar as espécies e seus respectivos estádios foram de 4.500 a 6.000 cm-1 e de 4.500 a 6.000 + 6.500 a 7.750 cm-1, com assertividade de até 100% para maioria das classes, sem a necessidade de pré-processamento. Os resultados destes estudos demonstraram que a espectroscopia NIR pode ser utilizada para discriminar espécies de corda-de-viola em diferentes estádios de crescimento, representando um avanço na pesquisa e implantação desta técnica para o desenvolvimento de equipamentos que auxiliem a adoção e/ou realização de um manejo específico de plantas daninhas.
The occurrence of weeds in cultivated areas is one of the main factors that limit agricultural productivity and increase production costs. The specific management of these species has gained prominence in recent years, mainly due to the advancement of technologies and for presenting an alternative to reduce the use of herbicides and environmental exposure. The great challenge for specific weed management is to identify the target species among the cultivated plants. In this context, several technologies have been studied for species identification and field recognition. Near-infrared spectroscopy (NIRS) is widely used in the analysis of quality and characterization of food and soils, but despite being a widespread technique, there are still few studies related to its use for weed identification/discrimination. The objective of this work was to verify the discriminatory potential of NIRS associated with multivariate analyzes in species of morning glory and growth stages. Thus, experiments were carried out with the following objectives: (I) discriminating the species of Ipomoea hederifolia, Ipomoea nil, and Merremia aegyptia, using multivariate techniques and pre-processing and spectral range selection; (II) discriminate the species of I. hederifolia and M. aegyptia, considering three growth stages.The sowing of the species was carried out in plastic pots with a capacity of 0.5 L, filled with sieved river sand, previously washed. The species were grown in an open environment and irrigated daily with water and nutrient solution. On the day of spectra collection, the plants were characterized for water content and stage (in the second experiment). The NIR spectra were acquired in absorbance in the spectral range of 10,000 to 4,000 cm-1 (1,000 to 2,500 nm). At the time of collection of the NIR spectra, the ambient temperature was maintained at 23 ° C ± 1 and the reading of the spectra was performed on the adaxial part of the fresh leaves, positioned on a black background. The spectral data were analyzed using the Unscrambler ® program. The pre-processing and spectral bands were tested and the spectral data submitted to the principal component analysis (PCA); linear discriminant of principal components (PC-LDA) and only in the first experiment, the discriminant analysis of partial least squares (PLS-DA). The study of the spectral ranges as well as the pre-processing were relevant and promoted improvements in the accuracy of the models. The accuracy values in the first experiment were up to 99.26, 98.52, and 98.73%, for I. hederifolia, I. nil and M. aegyptia, respectively. In the second experiment, the best intervals to discriminate species and their respective stages were 4,500 to 6,000 cm-1 and 4,500 to 6,000 + 6,500 to 7,750 cm-1, with assertiveness of up to 100% for most classes, without the need for pre-processing. The results of these studies demonstrated that NIRS spectroscopy can discriminate species of morning glory at different stages of growth, representing an advance in the research and implementation of this technique for the development of equipment that assists in the adoption of specific management.
The occurrence of weeds in cultivated areas is one of the main factors that limit agricultural productivity and increase production costs. The specific management of these species has gained prominence in recent years, mainly due to the advancement of technologies and for presenting an alternative to reduce the use of herbicides and environmental exposure. The great challenge for specific weed management is to identify the target species among the cultivated plants. In this context, several technologies have been studied for species identification and field recognition. Near-infrared spectroscopy (NIRS) is widely used in the analysis of quality and characterization of food and soils, but despite being a widespread technique, there are still few studies related to its use for weed identification/discrimination. The objective of this work was to verify the discriminatory potential of NIRS associated with multivariate analyzes in species of morning glory and growth stages. Thus, experiments were carried out with the following objectives: (I) discriminating the species of Ipomoea hederifolia, Ipomoea nil, and Merremia aegyptia, using multivariate techniques and pre-processing and spectral range selection; (II) discriminate the species of I. hederifolia and M. aegyptia, considering three growth stages.The sowing of the species was carried out in plastic pots with a capacity of 0.5 L, filled with sieved river sand, previously washed. The species were grown in an open environment and irrigated daily with water and nutrient solution. On the day of spectra collection, the plants were characterized for water content and stage (in the second experiment). The NIR spectra were acquired in absorbance in the spectral range of 10,000 to 4,000 cm-1 (1,000 to 2,500 nm). At the time of collection of the NIR spectra, the ambient temperature was maintained at 23 ° C ± 1 and the reading of the spectra was performed on the adaxial part of the fresh leaves, positioned on a black background. The spectral data were analyzed using the Unscrambler ® program. The pre-processing and spectral bands were tested and the spectral data submitted to the principal component analysis (PCA); linear discriminant of principal components (PC-LDA) and only in the first experiment, the discriminant analysis of partial least squares (PLS-DA). The study of the spectral ranges as well as the pre-processing were relevant and promoted improvements in the accuracy of the models. The accuracy values in the first experiment were up to 99.26, 98.52, and 98.73%, for I. hederifolia, I. nil and M. aegyptia, respectively. In the second experiment, the best intervals to discriminate species and their respective stages were 4,500 to 6,000 cm-1 and 4,500 to 6,000 + 6,500 to 7,750 cm-1, with assertiveness of up to 100% for most classes, without the need for pre-processing. The results of these studies demonstrated that NIRS spectroscopy can discriminate species of morning glory at different stages of growth, representing an advance in the research and implementation of this technique for the development of equipment that assists in the adoption of specific management.
Descrição
Palavras-chave
Agronomia, Ervas daninhas, Convolvulaceae, Quimiometria, Análise espectral, Agricultura de precisão