Development and Validation of Prediction Models for Rice Surface Lipid Content and Color Parameters Using Near‐Infrared Spectroscopy: A Basis for Predicting Rice Degree of Milling

Abstract

ABSTRACTDegree of milling (DOM) of rice plays a key role in determining rice quality and value. Therefore, accurate, nondestructive, quick, and automated surface lipid content (SLC) measurement would be useful in a commercial milling environment. This study was undertaken to provide calibration models for commercial use to provide quick and accurate evaluation of milled rice SLC and Hunterlab color parameters (L,a,b) as indications of rice DOM. In all, 960 samples, including seven cultivars from seven southern United States locations, stored for 0, 1, 2, 3, and 6 months, were milled for four durations to obtain samples of varying DOM. The samples were used to develop calibration models of milled rice SLC and L,a,b values. Another sample set (n = 58) was commercially milled and used to validate the developed models. A DA 7200 diode array analyzer was used to scan milled rice samples in wavelength spectra of 950–1,650 nm. SLC and color parameters were measured using a Soxtec system and a HunterLab colorimeter, respectively. The partial least squares regression (PLS) method using the full near‐infrared spectra was used to develop prediction models for rice SLC and color parameters. Milled rice SLC was well fitted with a correlation of determination of predicted and measured values of (R2 = 0.934). Color parameters were also successfully fitted for L (R2 = 0.943), a (R2 = 0.870), and b (R2 = 0.855). Performance of the developed models to predict rice DOM was superior in predicting SLC and L,a,b values with R2 predicted and measured values of 0.958, 0.836, 0.924, and 0.661, respectively.