International Journal of Engineering and Technology

Near infrared spectroscopic (NIRS) techniques have shown promise as rapid and non-destructive tools to evaluate the internal quality attributes of fruits. The objective of this work were to develop a calibration model for prediction soluble solids content (SSC) and acidity in ‘Gedong Gincu’ mango using NIRS and to analyze the effects of different preprocessing techniques on the acuracy of the calibration model. The prediction models were developed by partial least square (PLS) regression. NIR reflectance spectra were measured at wavelength of 1000-2500 nm using NIRFlex N-500 fiber optic solid. The results show that for SSC, the best preprocessing method was smoothng 3 points (sa3) with PLS factor=15, r=0.82, SEC=0.5ºBrix, SEP=1.28ºBrix, CVc 5.8% and RPD=1.52. For acidity, the best preprocessing method was first derivative Savitzky-Golay-9 ponts (dg1) with PLS factor=3, r=0.74, SEC=0.01%, SEP=0.12%, CVc=38.1% and RPD=1.33. This findings showed the important role of preprocessing method in developing accurate models for the prediction of mango internal quality characteristics.