Calibration model optimization for rice cooking characteristics by near infrared reflectance spectroscopy (NIRS)

The aim of this study was to evaluate the potential of near-infrared reflectance spectroscopy (NIRS) as a rapid method to estimate the cooking characteristics of rice. A total of 586 samples from rice breeding lines from 1999 to 2002, which had high variation for agronomy, location and year, were scanned by NIRS for calibration optimization by chemometric methods. Two subsets of 212 samples from one year and 400 samples from three years were employed to find suitable sample status for extending the NIRS utilities. There were three algorithms of PCA, PL1 and PL2, in which the first one was only based on the sample spectra variant and the remaining two based on the variant of both spectra and chemical characteristics to describe the relationship between any two neighboring samples. According to the results of calibration and validation by the three algorithms used, the suitable calibration samples could be chosen by the cutoff of neighborhood distance (NH) of 0.35, 0.4 and 0.45, respectively. For the cooking characteristics, the combination of SNV+D/‘1,4,4,1’ was the best pretreatment and the accuracy models were obtained with low SECV and high 1 − VRof amylose content (1.42% and 0.95%), gel consistency (9.49 and 0.76 mm) and alkali spread value (0.86 and 0.79 grade) . The models developed using brown rice and milled rice were superior to those using intact rice grains, but slightly poorer to those using their corresponding flour samples. Therefore, on-line monitoring of rice quality could be conducted in rice processing at milling stages. Due to the fewer sample mass destroyed, the brown rice (3 g) and brown rice flour (3 g or 0.5 g), by which the NIRS models were successfully developed for cooking characteristic analyses, could be introduced into quality evaluation of germplasm and intermediate lines selection in breeding projects.