[Influence of improved Kennard/Stone algorithm on the calibration transfer in near-infrared spectroscopy]

The Kennard/Stone(K/S) algorithm is adopted to select the transfer set samples in the near infrared analysis calibration transfer by calculating the Euclidean distance of the spectrum of the samples. In order to get a best expression of the comparability of the samples and select the best samples for calibration transfer, piecewise direct standardization (PDS) algorithm was investigated for resolving calibration transfer of near-infrared spectra of mixed amine, then the Euclidean distance and the Mahalanobis distance were applied to calculate the comparability of the spectra of the samples and the standard error of prediction (SEP) was used to evaluate them. The combination of spectral difference and property difference was introduced to estimate the effect of individual spectral difference and property difference. The experimental results showed that the Mahalanobis distance was better than the Euclidean distance in PDS and the combination of spectral difference and property difference was more effective in expressing the comparability of the samples.