A comparative study of acoustic and ultrasonic nondestructive testing for evaluating melon quality

Melon (Cucumis melo L.) is a high-value agricultural commodity known worldwide for its sweet
taste and crisp flesh texture, which are important factors for quality and consumer acceptance.
Unfortunately, quality testing and determining the optimal harvest time for achieving desired melon
characteristics are traditionally performed through destructive methods. The aim of this study was to explore
the potential of acoustic and ultrasonic tests for predicting the physicochemical properties of Honey Globe
melons (Cucumis melo L. var. inodorus). A total of 100 melon samples were used in this study. For the
nondestructive ultrasonic testing, attenuation values served as its variable, whereas in acoustic testing,
variables included frequency, magnitude, short-term energy, and zero-moment. Fruit's flesh firmness and
total soluble solids (TSS) as physicochemical quality properties were determined using destructive tests.
The calibration phase involved 80 melon samples, employing a K-Fold Cross Validation approach with ten
folds, done on Partial Least Square Regression (PLS) modeling. Another 20 melon samples were used for
blind testing. Reliability evaluation was done on key metrics, consisting of R2 values, RMSEC (Root Mean
Square Error of Calibration), RMSECV (Root Mean Square Error of Cross- Validation), RMSEP (Root
Mean Square Error of Prediction), and RPD (Ratio of Performance to Deviation). Analysis results on these
metrics collectively support the conclusion that both ultrasonic and acoustic methods exhibit their potential
to predict the firmness properties of melon fruits. The bestevaluation result that has been conducted for the
ultrasonic test uses attenuation, age, and density as predictors to predict fruit firmness, with R 2 = 0.763 and
RPD = 2.945, while the acoustic test achieved the best result with magnitude used as a predictor to predict
fruit firmness with R 2 = 0.718 and RPD = 2.230. However, evaluation metrics on the prediction of total
soluble solids for both nondestructive tests were still not good enough for application with low R2 and RPD
value.