Abstract
The combination of modified optical path length estimation and correction (OPLECm) and moving window partial least squares (MWPLS) was applied to the waveband selection for the Fourier transform near-infrared (FT-NIR) determination of pectin in shaddock peel samples. For MWPLS modeling, the changeable parameter of waveband width was set as the 21 piece-window size selected by the piece-wise multiplicative scatter correction (PMSC) method. Further, OPLECm was implemented for multiplicative scattering correction, and the MWPLS models were optimized by comparing the predictive results of three different cases using OPLECm pretreatment. The optimum case was PLS modeling based on the waveband selected by a moving window and successively pretreated by OPLECm (MW + OPLECm + PLS). The selected waveband was 8573–7784 cm−1 with a width of 410. The optimal number of PLS components was 11, and the corresponding RMSEP, the RP, and RRMSEP were 0.4176 wt%, 0.9185, and 8.43%, respectively. This model was better than other models for MW + OPLECm + PLS. The optimally selected waveband also showed an appreciative result in the validation part. This indicates that MWPLS modeling combined with OPLECm provides an appreciable improvement in the FT-NIR quantitative analysis of pectin in shaddock peel. In addition, a method was proposed for the classification of calibration samples and the prediction samples based on the principles of vector normalization, and all the calibration models and the prediction results were obtained on this classification.
Acknowledgments
This work was supported by the Natural Scientific Foundation of China (11226219), Guangxi Key Laboratory of Spatial Information and Geomatics (1103108-08), the Natural Scientific Foundation of Guangxi (2012GXNSFBA053013), and the Scientific Research Project of Guangxi Education Office (201203YB085).