Geographic origins of rice determinated by chemometric analysis of NIR spectra

A nondestructive optical method for determining the geographic origins of rice was investigated. Average absorption spectra of rice for three different geographic origins were analyzed. Direct orthogonal signal correction (DOSC), standard normal variate transformation (SNV) combined with detrending, multiply scatter calibration (MSC) and Savitzky-Golay second-order derivative transformation (S.Golay 2nd-Der) were applied to pre-process the original spectral data. DOSC present best preprocessing methods in this study and was proved to be more effective than the classical pretreatment methods, such as SNV and MSC. Through a chemometrics method of partial least squares (PLS), a correlation was established between the absorbance spectra and geographic origins of rice. 7 principal components (PCs) were obtained by PLS as the best number of principal components, and finally 30 unknown rice samples were successfully predicted based on 7 PCs as the input of BP neural network. The accuracy of prediction of different geographic origins was 83%. Vis/NIRS technique combined with chemometrics was promising for distinguishing the geographic origins of rice fast and nondestructively. This new combination of DOSC with PLS and BPNN might be applied for other materials based on NIR spectroscopy.