Nondestructive Prediction of Juice Recovery Yield of Pineapple Using Near Infrared Hyperspectral Imaging

Abstract

During commercial processing of pineapples, fresh fruit selection on the basis of their quality is essential, particularly their juice content. This is to ensure high and consistent product quality, but juice level varies between individual fruit. Therefore, a non-destructive technique for predicting juice recovery yield of pineapple using near infrared hyperspectral imaging (NIR-HSI) was aimed for use in online sorting systems. Pineapples were scanned using NIR-HSI to develop a calibration model for predicting juice recovery yield of pineapple in this study. A set of 122 pineapple samples was divided into a calibration set (n = 81) and a prediction set (n= 41). Spectral pretreatments were investigated in order to obtain the best calibration model. The best model was obtained using Savitzky-Golay smoothing spectral pretreatment at the wavelength range of 935–1720 nm using partial least squares regression (PLSR). The model showed sufficient accuracy for prediction with a correlation coefficient (Rp) of 0.73 and the root mean square error of prediction (RMSEP) of 1.54%. These results indicate that NIR-HSI has the potential for use in prediction the juice recovery yield of pineapple in a non-destructive online system in pineapple processing factories.