Desenvolvimento de um sistema integrado de classificação colorimétrica para identificação de estágios de maturação de tomates tipo Roma

Abstract

The food industry is facing a growing demand for precise and reliable methods to assess the quality of tomatoes, with an emphasis on the color parameter, which is crucial for determining the ripeness stage and consumer acceptance. Traditional visual analysis, although widely used, presents significant deficiencies, such as subjectivity, resulting in variability among evaluators. In contrast, computer vision using optical sensors offers an objective and consistent approach, allowing for precise classification of ripening stages based on quantitative data. This study developed an automatic colorimetric classification method for Roma tomatoes at different ripening stages, utilizing the TCS3472 optical sensor integrated with Arduino. The research aimed to meet the food industry's need for precise, non-destructive methods capable of distinguishing between green, semi-ripe, and ripe tomatoes. Univariate and multivariate classification models were used, including principal component analysis (PCA), principal component discriminant analysis (PCA-DA), and partial least squares discriminant analysis (PLS-DA). The models' performance was evaluated in terms of accuracy, sensitivity, specificity, and correct classification rate (CCR), comparing the RGB and CIE Lab* color spaces along with reflectance. The results indicated that the RGB model, especially when used in SIMCA and PLS-DA, achieved a high CCR, above 97%, with 98% sensitivity for ripe tomatoes, 94% for semi-ripe, and 100% for green. The approach based on the difference between the blue (B) and red (R) channels proved particularly effective, surpassing models based on individual R and G signals. On the other hand, the CIE Lab* model performed better in identifying green tomatoes but showed difficulties in classifying ripe and semi-ripe tomatoes. The research concludes that the RGB-based method is more suitable for applications requiring high precision and reliability in classifying tomato ripening stages, while the CIE Lab* reflectance-based method may be useful in contexts where identifying green tomatoes is critical. The practical application of the proposed method can optimize tomato selection and processing in the industry, contributing to reducing post-harvest losses and increasing production process efficiency.