Prediction of Cadmium content in brown rice using near-infrared spectroscopy and regression modelling techniques

The feasibility of prediction of cadmium (Cd) content in brown rice was investigated by near-infrared spectroscopy (NIRS) and chemometrics techniques. Spectral pretreatment methods were discussed in detail. Synergy interval partial least squares (siPLS) algorithm was used to select the efficient combinations of spectral subintervals and wavenumbers during constructing the quantitative calibration model. The performance of the final model was evaluated by the use of root mean square error of cross-validation (RMSECV), root mean square error of prediction (RMSEP) and correlation coefficients for calibration set and prediction set (R-c and R-p), respectively. The results showed that the optimum siPLS model was achieved when two spectral subinterval and fifty-two variables were selected. The predicted result of the best model obtained was as follows: RMSECV=0.232, R-c=0.930, RMSEP=0.250 and R-p=0.915. Compared with PLS and interval PLS models, siPLS model was slightly better than those methods. These results indicate that it is feasible to predict and screen Cd content in brown rice using NIRS.