Visible and Near-Infrared Spectroscopy for Determining Pomegranate Juice Volume

Pomegranate fruit is one of the essential garden products in Iran, which, in addition to its high economic-commercial value, has many nutritional benefits. Due to the lack of appropriate development of the industry in the processing sector for this product, factories use completely traditional and limited methods for measuring the quality of pomegranate, which reduces the quality of the processed product. In the present study, the feasibility of visible and near-infrared (Vis/NIR) spectroscopy as non-destructive technology was used to determine the pomegranate fruit juice volume, which is one of the crucial quality attributes of the product. Spectral data of the samples obtained by applying waves in the range of 400 to 2500 nm, were evaluated by five methods including multiplicative scatter correlation, standard normal variate (SNV), vector normalization, first derivative, second derivative, as well as non-processed state. Then partial least square regression (PLSR) was used to estimate the juice volume. These processes were implemented according to machine learning algorithms using PYTHON 3.8 software. Lowest modeling error (18.61) was achieved with 36 principal components of extracted features from spectral signature. The results showed that this method estimated the amount of pomegranate juice with a coefficient of determination of 94 %, mean absolute error of 5.3 and distance of 0.98. This outcome resulted from combination of SNV preprocessing with PLSR regression in the range of 400 to 2500 nm. In spite of the effect of preprocessing for juice volume estimation, Vis-NIR spectroscopy possess desired condition for estimation of pomegranate juice volume.