Water content prediction of ‘crystal’ guava using visible-near infrared spectroscopy and chemometrics approach

Abstract

The purpose of the research was to apply the visible-near infrared (vis-nir) spectroscopy and chemometrics approach to predict the water content of crystal guava rapidly and non-destructively. The absorbance spectra were collected from intact ‘crystal’ guava fruits in wavelength from 381 to 1065 nm. Spectra pre-treatment was used to enhance the prediction accuracy. Several spectra pre-treatment methods were arranged to the original absorbance spectra such as multiplicative scatter correction (MSC), standard normal variate (SNV), second derivative absorbance (d2a), MSC+d2a, and SNV+d2a. Calibration model was developed by partial least squares regression (PLSR). Validation was done by K-fold cross-validation. The results presented that vis-nir spectroscopy combined with chemometrics approach gave accurate water content prediction of ‘crystal’ guava. The best calibration model was provided by spectra pre-treatment of SNV+d2a with coefficient of determination (R2) and ratio performance deviation (RPD) of 0.74 and 1.60, respectively. It concluded that vis-nir spectroscopy and chemometrics approach can be performed to predict the water content of ‘crystal’ guava rapidly and non-destructively by replacing the standard laboratory analysis.