Amylose content calibration model for the three types of selected rice grains using visible shortwave near infrared spectroscopy

Abstract

Amylose content is one of the main characteristics to measure the quality and texture of rice. This research aims to conduct a non-invasive measurementof amylose content in rice grains using a Visible-Shortwave Near-Infrared Spectroscopy (VISSWNIRS) through the combination of two methods: Principal Component Regression (PCR) and Artificial Neural Network (ANN). Three data sets of rice samples (spectral VIS-SWNIR and amylose content reference) from three types of rice (brown rice, basmati rice and white rice) that are available in the Malaysian market were used and processed separately. The effect of data shift in the reflection spectrum was eliminated using the zero, first and second order derivatives which were then combined with the zero, first and second order of the Savitzky-Golay filter. The data spectrum spread was reduced using Singular Value Decomposition (SVD). The PCR and ANN methods were applied with 65% of the data sets were used for training while the remaining 35% were used for testing. The research analysis results have found that the Root-Mean-Square-Error of Calibration (RMSEC),the correlation coefficient of calibration (rc), the Root-Mean-Square-Error of Prediction (RMSEP), and the prediction correlation coefficient (rp) of PCR for brown rice were 2.96, 0.44, 2.74, and 0.22 respectively. For basmati rice, the corresponding values were 1.93, 0.57, 1.98, and 0.40 while for white rice the values were 2.42, 0.73, 2.65, and 0.62. In the meantime, ANN analysis yields the values of 0.70, 0.99, 0.96, and 0.88 for brown rice, 0.24, 0.99, 0.31, and 0.99 for basmati rice and 1.03, 0.95, 1.05, and 0.93 for white rice. The results suggest that VIS-SWNIRS is suitable and has the potential to be used in the non-invasive assessment of amylose content in rice grains from three types of rice in the Malaysian market.