Comparative Assessment of Transmission and Reflection Vis/NIR Spectroscopy for Non-Destructive Evaluation of Multiple Quality Attributes in Tomatoes

Abstract Non-destructive evaluation of internal and external quality attributes is imperative for effectively grading and sorting tomatoes. This study compared visible/near-infrared (Vis/NIR) diffuse reflectance and transmission spectroscopy for rapid, non-invasive measurement of key indicators, including color, hardness, total sugar (TS), and total acidity (TA). A sample set of 110 tomatoes across multiple ripeness levels was divided into calibration (n = 82) and prediction (n = 28) subsets. Vis/NIR spectra were obtained using reflectance and transmission systems and pre-processed before multivariate analysis. Partial least squares regression (PLSR) models were developed, relating the spectra to reference measurements using competitive adaptive reweighted sampling (CARS-PLS). For internal parameters of TS and TA, transmission PLS models demonstrated superior performance over reflectance, with prediction R values of 0.9511 and 0.9818. In contrast, for external attributes of color and hardness, reflectance PLS models performed better given consistent bulk fruit maturity, with prediction R values of 0.9595 and 0.9713. This study demonstrates the potential of Vis/NIR diffuse transmission spectroscopy for non-invasive analysis of internal and external tomato quality attributes. The findings provide a basis for developing handheld devices and inline online systems for sorting tomatoes based on comprehensive ripeness assessment.