Estimativa de índices físico-químicos e sensoriais de frutas usando imagens digitais

Abstract

An increase in fruit consumption has been observed in the last decade, which demonstrates that the population has sought healthier lifestyles. Nevertheless, since consumers look for quality products, they often seek to evaluate visual aspects such as color and freshness. The determination of quality parameters in fruits is normally carried out using conventional analyses, which have disadvantages such as residue production and samples destruction. In recent years, smartphones have been used as an instrument in chemical analysis. Therefore, this study aimed to build calibration models based on digital images obtained with a smartphone to estimate sensory and physicochemical characteristics of bananas and papayas at different stages of maturation. Banana and papaya samples were analyzed at different stages of maturation. Three approaches were used to obtain image information: (i) average RGB values of the entire fruit image, (ii) number of pixels in each RGB value (RGB profile) and (iii) percentages of different colors exhibited by the fruit during ripening. Predictive models were built to predict sensory parameters of ideal sweetness, ideal firmness and global acceptance, as well as SST and firmness instruments. The models were calibrated by multiple linear regression (MLR) and partial least squares regression (PLS). The approach based on average RGB values showed better performance to predict sensory and instrumentally analyzed parameters. Thus, the use of images obtained by smartphone has great potential to estimate the quality parameters that most influence acceptance by consumers.