A Novel Calibration Transfer Method of Near Infrared Spectral Model Without Standard Samples Based on Screening Stable and Consistent Wavelengths

The change of measurement environment and spectral signals of spectroscopy instruments can lead to big error in the process of transferring the model of near infrared (NIR) spectra to slave instruments. Present work advanced a novel method of transferring NIR calibration models without standard samples basing on Screening stable and consistent wavelengths (SSCW). The method firstly eliminates the wavelengths at which the standard variation of difference spectra between master and slave instruments (SDDSI) is bigger than the standard variance of precision detection spectra (SDPDS) to screen the wavelengths at which the spectral signals of different instruments are consistent well. And then deletes the wavelengths at which the values of SDPDS are too high. Thus the spectral signals at the selected wavelengths are stable and consistent between instruments to build the NIR calibration models. Two datasets are applied to validate the method of SSCW : one is consisted of 72 radix scutellaria samples, whose NIR spectra are measured on two different types of NIR spectroscopy instruments, another one is consisted of 80 corn samples, whose NIR spectra measured on three NIR instruments are available online. The results show that the overall prediction effect of SSCW for salves' samples is much better than that of the full-wavelength PLS model. In most cases, the root mean square of error prediction (RMSEP) of salve samples of SSCW is lower than those given by the piecewise direct standardization (PDS) correction, and that of the other model transfer methods without standard samples. Therefore, the method of SSCW is of fewer parameters, robust and can maintain good prediction performance after transferring to salve instruments. It can be used to realize the sharing of PLS calibration model among spectroscopy instruments.