Prediction models for degree of milling (DOM) of rice (Oryza sativa L.) using near infrared spectroscopy
2017
Omega, C. A.
A non-destructive and rapid technique near infrared (NIR) spectroscopy was applied for the prediction of the degree of milling (DOM) of rice. Sixteen rice cultivars harvested from 2015 to 2016 were milled for 20s, 40s, 60s, 80s, and 100s to obtain various milling degrees for building calibration models and testing validation sets. Freshly husked brown rice was included as an unmilled treatment validation sets. Freshly husked brown rice was included as an unmilled treatment. A NIRQuest512-1.7 spectrometer with an actual wavelength range of 900-1700nm was utilized for acquiring spectral data of milled rice. The performed mathematical manipulations for the spectra were first derivative, second derivative, wavelet detrending, multiple scatter corrections, and mean centre to maximize extractions of relevant information. The partial least square regression (PLSR) method was applied for optimizing wavelength range by pre-processing and pre-treatments technique. Relatively good models were predicted for whiteness, gravimetric, surface lipid content (SLC), and alcohol-alkali staining method with an adjusted R2 of 0.971, 0.873, 0.95, and 0.86, respectively. Results indicated that near infrared spectroscopy can perform well predicting the desired parameters when subjected for testing through rough screening up to quality control. A good prediction model for SLC was acquired on the second overtone and combination bands of C-H functional groups present in fatty acids and oils.
显示更多 [+] 显示较少 [-]