Combined use of a near-infrared spectrometer and a visible light grain segregator for accurate non-destructive determination of amylose content in rice

Amylose content contributes to the texture and the eating quality of rice. Industrial non-destructive measurement of the amylose content in rice using a near-infrared (NIR) spectrometer remains inaccurate. In this study, we analyzed whether using an NIR spectrometer along with a visible light (VIS) grain segregator improved the accuracy of conventional model for non-destructive determination of rice amylose content. NIR spectra collected from the NIR spectrometer and physical properties were collected from the VIS grain segregator from 603 milled samples of 10 varieties of rice produced between 2010 and 2016 in various regions of Hokkaido, Japan. Data were analyzed by partial least squares regression (PLS) and multiple linear regression (MLR) to develop a dual-step model. Our dual-step model was validated using production year samples (2016) that were different from the calibration set (2010-2015). Our method improved accuracy compared to the conventional method developed using data from an NIR spectrometer, with a standard error of prediction (SEP) = 0.94% and ratio of performance deviation (RPD) = 2.46. Our dual-step model also had the highest robustness when increasing the production year samples of the calibration set, enabling a more precise, accurate, and efficient rice quality screening in Japanese industry.