Accurate Estimation of Ethanol Content in Fruit Juices using CIELab Color Space and Chemometrics via Smartphone-based Digital Image Colorimetry

Abstract

This study aims to investigate the optimal color space and chemometric technique for digital image colorimetry to determine ethanol content (% v/v) in apple, orange, and grape juices, using potassium dichromate (K2Cr2O7) under acidic conditions. The accuracy of colorimetric–chemometric integration across various color spaces (RGB, HSV, CIELab, CMYK, CIELuv, CIEXYZ, and CIELch) was benchmarked against UV–Vis spectrophotometry using metrics such as coefficient of determination (R²), mean absolute percentage error (MAPE), and root–mean–squared error (RMSE). Various chemometric techniques (PLS, PCR, MLR, multivariable–SVR, and multivariable NN regression) were evaluated. Results demonstrate that combining the CIELab color space with PLS or MLR yields the most accurate ethanol determination. Both techniques achieved average MAPE and %RMSE values below 10% (7.026% and 7.78% for PLS; 7.34% and 7.94% for MLR) and a competitive limit of detection of 0.02% (v/v) at best, and 0.087% (v/v) on average, indicating excellent model predictability and accuracy.