A green method for the prediction of color in sugar using fingerprint signatures based on digital images and chemometrics

This paper proposes a new analytical methodology for color determination in sugar samples according to the ICUMSA (International Commission for Uniform Methods of Sugar Analysis) standard, one of the main quality parameters of this food. To this purpose, digital images and multivariate calibration will be associated. The sugar samples were digitally captured using a flatbed scanner, and histograms were generated to depict the frequency distribution of color indices in the gray scale (Gray), Red-Green-Blue (RGB), and Hue-Saturation-Intensity (HSI) channels, as well as their combinations. Multivariate calibration models were developed and compared, employing Partial Least Squares (PLS), Successive Projections Algorithm associated with Multiple Linear Regression (SPA-MLR), and Genetic Algorithm associated with Multiple Linear Regression (GA-MLR). Better results were achieved with the SPA-MLR model, exhibiting RMSECV, RMSEP, REP, r 2 pred, and RPDpred values of 9.5 IU, 4.9 IU, 9.879 %, 0.976, and 6.289, respectively. These results align with the reproducibility criterion of the reference method, highlighting the advantages of the proposed method, which is simple, rapid, cost-effective, and eliminates the need for chemical reagents.