Open AccessPrediction of surface color of ‘crystal’ guava using UV-Vis-NIR spectroscopy and multivariate analysis
Author(s) -
Kusumiyati Kusumiyati,
Wawan Sutari,
Farida Farida,
Syariful Mubarok,
Jajang Sauman Hamdani
Publication year - 2019
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/365/1/012026
Subject(s) - partial least squares regression , near infrared spectroscopy , ultraviolet visible spectroscopy , spectroscopy , ultraviolet , wavelength , analytical chemistry (journal) , multivariate statistics , calibration , materials science , mathematics , chemistry , optics , chromatography , optoelectronics , statistics , physics , organic chemistry , quantum mechanics
Ultraviolet, visible, and near infrared (UV-Vis-NIR) spectroscopy technique has been increasingly applied in sorting of agricultural products, especially fruits. This study aimed to evaluate the use of UV-Vis-NIR spectroscopy and multivariate analysis to predict the surface color (L * , a * , and b * ) of ‘crystal’ guava. Wavelengths covering the UV-Vis (300-699 nm) and UV-Vis-NIR (300-1065 nm) area were investigated to obtain the best prediction accuracy. A total of 120 samples of guava fruits were harvested at the same maturity level, the samples were divided into 3 groups, which consisted of 40 samples per group. The first group was stored for 0 days. The second and third groups were stored for 4 and 8 days, respectively. Spectra data acquisition was performed at wavelengths of 300-1065 nm and interval of 3 nm. Multiplicative scatter correction (MSC) and standard normal variate (SNV) spectra correction methods were applied to improve prediction accuracy. Partial least squares regression (PLSR) was used as calibration method. Validation was done by the k fold cross validation. The best prediction of L * was obtained at UV-Vis-NIR wavelength with SNV correction method. The UV-Vis wavelength and SNV acquired the best prediction of a * . Original spectra and UV-Vis-NIR wavelengths resulted the best prediction of b * .