Abstract: Based on NIRS technique and partial least squares (PLS) algorithm, two calibration models were established to analyze quantitatively fat content (FT) in brown rice grain and brown rice flour, respectively. The determination coefficients (R2) of these two models for FT were 94.44% and 95.54%, and the root mean square errors of cross validation (RMSECV) were 0.09% and 0.08%, respectively. In external validation, the R2 value between the true value and predicted value were 79.51% and 87.10% for FT in brown rice grain and brown rice flour, and the root mean square errors of prediction (RMSEP) were 0.24% and 0.26%, and the average relative errors were 4.11% and 3.30%, respectively. These results indicated that the method of NIRS has relatively high accuracy in prediction tests for FT in brown rice grain and brown rice flour and that the two mathematic models established in the present study should be useful for nutrient quality improvement in rice breeding program.

Key words: Rice, Fat content, NIRS (near infrared spectroscopy), Quantitative analysis