Determinação do teor de clorofilas e carotenoides em azeites de oliva e abacate usando um smartphone

Abstract

Olive oils are known for their high nutritional value. Its pigments, chlorophylls, and carotenoids, have properties that go beyond their function in coloring, bringing health benefits as bioactive compounds. The determination of these pigments in olive oils is usually carried out using spectrophotometric methods, which use expensive equipment and which do not allow analysis outside the laboratory. On the other hand, smartphone cameras have shown great potential as instruments for chemical analysis. Thus, this work aimed to evaluate the potential of using digital images obtained with a smartphone and multivariate calibration to determine the content of chlorophylls and carotenoids in olive oils. Images of 46 samples of olive oil and avocado, placed inside a glass cuvette, were obtained using a smartphone. The data were obtained in triplicate, totalizing 138 observations. A test set, external to the calibration, was made up of 25% of the observations. Color components in the RGB, Y, HSV, CMYK, Lab, or XYZ systems of the olive oil images were calibrated using Multiple Linear Regression (MLR) and Least Squares Support Vector Machines (LS-SVM) against the respective levels of chlorophylls and carotenoids obtained by the spectrophotometric method. A type of calibration transfer based on images of a FeCl3.6H2O solution was proposed, evaluating its application under different analytical conditions, such as different smartphones and lighting. A good performance of the models was observed for both pigments, and the models built using LS-SVM were superior to those obtained by MLR. The type of lighting influenced the prediction errors of the models more than the type of smartphone. However, considerable improvements were obtained when applying the calibration transfer proposal, decreasing the RMSE and increasing the R² for both chlorophylls and carotenoids. The best performance after applying the calibration transfer was obtained using the XYZ/LS-SVM model for chlorophylls and Y/LS-SVM for carotenoids, which had an average RMSE of 0.96 and 0.29 mg.Kg-1 with an average R² 0.96 and 0.83 for chlorophylls and carotenoids respectively. Therefore, the method based on digital images obtained by smartphone and the proposed calibration transfer showed great potential in the determination of pigments in vegetable oils, with good precision under different types of lighting or smartphone.