Implementing near infrared spectroscopy for the online internal quality and maturity stage classification of intact watermelons at industry level

Abstract

The industrial implementation of non-destructive techniques for the classification of watermelons, according to their quality standards and stage of maturity, is highly sought by the handling and processing industry. This study aimed to evaluate the feasibility of near-infrared spectroscopy (NIRS) for the individual internal quality assessment of intact watermelons, simulating industrial sorting lines. Two online near infrared (NIR) sensors, a diode array (DA) and a Fourier-transform (FT) spectrometer, each characterised by distinct optical configurations and technical specifications, were utilised. These sensors operated in reflectance mode, analysing the fruits in both static mode (conveyor belt stopped) and dynamic mode on a moving conveyor belt at two different speeds. Regression and classification models were developed for the prediction of soluble solid content (SSC) and the classification of the maturity stage, respectively, by applying various signal pre-treatment methods to the NIR spectra. The best results for SSC prediction were achieved using the DA instrument in dynamic mode, with no significant differences (P > 0.05) between the two conveyor speeds tested. Specifically, a residual predictive deviation for cross-validation (RPDcv) of 1.41 was achieved with the DA sensor in dynamic mode and a conveyor speed of 10.5 cm s−1. Furthermore, for the same instrument, mode, and speed, the proportion of fruits accurately classified as ‘mature’ and ‘immature’ in the training set was 76 % and 82 %, respectively, with corresponding values of 90 % and 70 % for the validation set. The findings are promising for the horticultural industry, demonstrating the potential for incorporating NIRS technology into industrial sorting lines for the internal quality assessment of individual watermelons.