Metodologia baseada em imagem digital, espectros UV-Vis e quimiometria para screening de adulteração de café por cascas e paus

Abstract

The quality of the coffee depends on several factors related to all stages of production, from the choice of species (or varieties) and from the transformations during the roasting to the preparation of the beverage. However, quality can be altered by the illicit practice of adulteration, which consists of contamination by husks and sticks and other adulterants to roasted and ground coffee. Thus, tampering causes the alteration of sensory properties and causes costly damage to the soil due to the high coffee consumption by Brazilians. In view of the above, the following work proposes the development of new, simple, fast and low-cost methodologies combining the use of digital images and ultraviolet and visible molecular absorption spectroscopy (UV-Vis) with the methods of recognition of supervised standards such as SIMCA, PLS-DA and SPA-LDA, for the construction of chemometric classification models. In fact, the constructed models discriminated and classified the adulterated coffees of the non-adulterated coffees, which can be verified through the performance evaluation parameters of the models, related to accuracy, sensitivity and specificity in the training and test sets presented in Table 1.1 and Table 1.2. Therefore, the best results were obtained for the digital and ultraviolet (UV) images with the PLS-DA and SPA-LDA modeling compared to SIMCA modeling. In the modeling of the digital images, the PLS-DA presented the following percentage of correctness for the training and test sets, according to Table 1.1: PLS-DA accuracy (73.0%), sensitivity (69.0%) and specificity (76.0%); accuracy (90.0%), sensitivity (94.0%) and specificity (85.0%) and SPA-LDA ; accuracy (95.0%), sensitivity (95.0%) and specificity (95.0%); SIMCA accuracy (60.0%), sensitivity (80.0%) and specificity (51.1%); accuracy (82.5%), sensitivity (93.3%) and specificity (76.0%). In the modeling of the ultraviolet (UV) aqueous extracts the PLS-DA, SPA-LDA and SIMCA presented the following percentage of correct answers for the training and test sets according to Table 1.2. PLS-DA accuracy (96.8%), sensitivity (97.3%) and specificity (96.1%); accuracy (97.4%), sensitivity (100.0%) and specificity (94.7%); SPA-LDA accuracy (100,0%), sensitivity (100,0%) and specificity (100,0%); accuracy (100.0%), sensitivity (100.0%) and specificity (100.0%); SIMCA accuracy (73.0%), sensitivity (62.2%) and specificity (88.5%); accuracy (79.5%), sensitivity (65.0%) and specificity (94.7%). Therefore, the proposed methodologies developed can ensure the purchase of coffee marketed without the presence of the adulterant, husks and sticks. In this way, it will be possible to mitigate costly damages to society and help the quality control of the coffee to the inspection and commercialization (MAPA, ANVISA and ABIC).