Monitoring postharvest water loss in eggplants (Solanum melongena L.) using UV-induced fluorescence imaging and multivariate analysis

Eggplant ( Solanum melongena L. ) is susceptible to significant postharvest losses primarily due to water loss during storage, which affects market quality by causing texture and glossiness degradation. We investigated whether UV-induced fluorescence imaging and EEM (Excitation-Emission Matrix) fluorescence spectroscopy can non-destructively monitor WL under four storage regimes (10 °C/95 % RH, 20 °C/95 % RH, 20 °C/75 % RH, 10 °C/75 % RH). EEMs exhibited three regions; a 365/420 nm blue emission increased most under warm, low-humidity storage and is consistent with phenolic/lignin-related fluorescence. Side-view fluorescence (FL) images showed progressive blue-white emission and surface textural changes that tracked gravimetric water loss (WL). A PLSR model using combined color and texture features from FL and reflectance (CL) images achieved R 2 CV = 0.88 (RMSECV = 3.47 %) with only six features. To test a minimal predictor, we fit an Analysis of Covariance (ANCOVA) using Day-1 FL MeanBlue as a covariate and storage category as a factor with Leave One Out Cross-validation (LOOCV); this forecasted cumulative WL with R 2 LOOCV = 0.92 and MAE = 1.88 %. Importantly, this ANCOVA model using Day-1 blue-band fluorescence as a covariate was predictive only under 20 °C/75 % RH; under the other conditions, its contribution was weak. Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) models achieved accuracies of 94.4 % and 85.2 %, respectively, in differentiating storage conditions. These results support low-cost FL imaging as a practical tool to monitor WL and storage stress. • UV-induced fluorescence imaging tracks eggplant water loss in storage. • Blue-band fluorescence (365/420 nm) rises under warm, dry storage. • Six-feature PLSR estimates WL% (R 2 CV = 0.88; RMSECV = 3.47 %). • Day-1 FL_MeanBlue + ANCOVA predicts Day-10 WL% (R 2 LOOCV = 0.92). • LDA and SVM classified storage conditions with up to 93.6 % accuracy.