The best calibration model for SLC was developed using PLS
regression based on the FT-NIR spectra processed by the FDþ MSC
treatment in the optimal wavelength ranges of 7501.7e5449.8 þ
4601.3e4246.5 cm1. FT-NIR spectroscopy analysis appeared to be
a rapid and precise technique to measure SLC of milled rice. SLCwas
found to decrease markedly as the milled rice grade was reduced.
The relationship between SLC and the grades could be fitted with
the linear function or the logarithmic model. However, the milled
rice grades could not be accurately calculated with the linear or the
logarithmic models. A three layer of BP-ANN model for milled rice
grades was developed based on the SLC predicted by NIR spectroscopy.
Conducting the BP-ANN model on the independent
testing set resulted in a CCP of 95.45%, indicating that the proposed
NIR method was a potent alternative to the traditional visual
inspection method for milled rice grade.