Kepok banana TSS and firmness estimation based on RGB reflectance–fluorescence imaging and partial least squares regression

This study presents a concise and improved approach for estimating the ripeness parameters of Kepok bananas ( Musa balbisiana BBB) using a Partial Least Squares Regression (PLSR) method. The proposed technique integrates RGB reflectance and fluorescence maging to obtain comprehensive color and texture information for objective fruit maturity estimation, specifically to determine the optimum harvest maturity. Comparative models were developed using individual and combined image datasets to evaluate their performance in predicting Total Soluble Solids (TSS) and firmness, two key indicators of banana maturity. Results revealed that fluorescence and combined RGB reflectance–fluorescence imaging yielded the highest accuracy, with R²training = 0.9231, R²test = 0.9256 for firmness and R²training = 0.8623, R²test = 0.8908 for TSS. The optimal PLSR models utilized 12 and 9 selected features, respectively, achieving RMSE values of 0.3341 (firmness) and 2.0829 (TSS). These outcomes confirm that integrating RGB reflectance–fluorescence imaging with PLSR modeling enhances the accuracy and reliability of non-destructive Kepok banana maturity assessment.