Use este identificador para citar ou linkar para este item: http://repositorio.ufla.br/jspui/handle/1/59223
Título: Detection of blends and prediction of natural pigments in vegetable oils using a low-cost color sensor
Título(s) alternativo(s): Detecção de misturas e predição de pigmentos naturais em óleos vegetais usando um sensor de cor de baixo custo
Autores: Nunes, Cleiton Antonio
Papaioannou, Emmanouil H.
Nunes, Cleiton Antonio
Freitas, Matheus Puggina de
Pedroso, Marcio Pozzobon
Rocha, Roney Alves da
Silva, Luiz Fernando de Oliveira da
Palavras-chave: Clorofilas
Carotenoides
Quimiometria
Colorimetria
Chlorophylls
Carotenoids
Chemometric
Colorimetry
Data do documento: 16-Ago-2024
Editor: Universidade Federal de Lavras
Citação: LORENZO, N. D. Detection of blends and prediction of natural pigments in vegetable oils using a low-cost color sensor. 2024. 66 p. Tese (Doutorado em Agroquímica) - Universidade Federal de Lavras, Lavras, 2024.
Resumo: In this study, the potential of using a color sensor to identify different mixtures of vegetable oils and predict levels of natural pigments in avocado oil was evaluated. The objectives were to authenticate avocado oil against blends containing canola, sunflower, corn, olive and soybean oils, as well as to explore the use of color sensing to predict total carotenoid and chlorophyll content in avocado and olive oils, in addition to their total spectrophotometric color (TSC). The study involved RGB interference under two types of lighting: UV (395 nm) and white light. Classification methods such as Linear Discriminant Analysis (LDA) and Least Squares Support Vector Machine (LS-SVM), as well as Multiple Linear Regression (MLR) and LS-SVM were used for detection and quantification of mixtures. To predict total pigment levels, MLR and LS-SVM were employed using different color parameters (RGB, HSV or L*a*b*). In general, the LS-SVM models outperformed those of MLR and LDA, demonstrating good prediction capacity without evidence of random adjustments, both for the study involving mixtures and for those involving pigments. In the study with mixtures, the LS-SVM model achieved R² greater than 0.9 in all cases, including external validation. Furthermore, using white light, the LS-SVM models produced root mean square error (RMSE) values varying between 1.17 and 3.07%, indicating good accuracy in quantifying mixtures. Such results were demonstrated by the color sensor in identifying mixtures of other vegetable oils in avocado oil, showing its potential as an efficient and low-cost alternative for vegetable oil authentication. In the study involving pigment content, different methods were used to determine the pigment content used as a reference for building the models. UV illumination has been shown to improve the predictive performance of total chlorophyll content when the response is based on solvent-free methods (IUPAC and AOCS methods). In other cases, white lighting proved more effective. Color data in the HSV system were more effective in predicting total chlorophylls when referenced to solvent-free methods (R² > 0.9 for external validation). RGB data is the most appropriate way to predict TSC and total chlorophylls referenced in methods that use solvent dilution (R² > 0.9 for external validation). Finally, total carotenoid content was best predicted using L*a*b* values as descriptors (R² of 0.8 for external validation). Therefore, low-cost methods based on a color sensor have proven to be a promising alternative for predicting pigments and TSC in vegetable oils, without the need for sample pretreatment or dilution in toxic solvents.
Descrição: Arquivo retido, a pedido da autora, até dezembro de 2024.
URI: http://repositorio.ufla.br/jspui/handle/1/59223
Aparece nas coleções:Agroquímica - Doutorado (Teses)

Arquivos associados a este item:
Não existem arquivos associados a este item.


Este item está licenciada sob uma Licença Creative Commons Creative Commons