A New Spectra Transfer Method for Multivariate Calibration Model of Molecular Spectroscopy Analysis

Whether it is for near-infrared (NIR) spectra or mid-infrared (MIR) spectra, the issue of multivariate calibration model transfer between spectrometers has not been satisfactorily solved yet. To realize the sharing of model or spectra between the spectrometers, an improved piecewise direct standardization (PDS) method, Spectra-Angle-PDS (SA-PDS), is proposed in this work. Spectral angle (SA) measurement which can evaluate the similarity between spectra or the difference between the spectral vector and the shape of the spectral curve is used as a criterion for optimization of parameters of PDS. The model transfer no longer needs reference data so that the model transfer is not affected by the error of properties of validation samples and the model. The model transfer from master spectrometer to slave one or opposite on the contrary can be more easily realized with SA-PDS. The proposed spectra transfer method (SA-PDS) has been applied to predict the content of glycosides in tobacco and the content of wax in asphalt using NIR and MIR spectra respectively. The PDS method using RMSEP to select parameters has also been performed for comparative study of the proposed model transfer technique. For the spectra transfer from salve to master spectrometers of NIR, comparative experiment results have shown that the root-mean-square error of prediction (RMSEP)with the proposed SA-PDS method is reduced from 5.257 4 (before transfer) to 1.337 1 (after transfer), and much smaller than that with PDS method (1.350 3). For the spectra transfer from salve to master spectrometers of MIR, the precision of model transfer also improved significantly, accompanied by the reduction in RMSEP from 0.525 1 to 0.186 9, is proven to be better than that with PDS method (0.219 4). The satisfactory spectra transfer results with the proposed SA-PDS method from the master spectrometer to the slave spectrometer has also been verified.