| Abstract: | A new calibration transfer method that applies canonical correlation analysis (CCA) to transfer the informative components extracted from a spectral dataset is proposed to reduce the interference of noise, background and non‐predicted properties. This method employs the partial least squares method to extract the informative components related to the predicted properties from the raw spectra and then corrects the informative components based on CCA. The performance of this algorithm was tested using three pairs of spectra batches: two pairs of corn spectra and one pair of tri‐component solvent spectra. The results showed that this method can significantly reduce prediction errors compared with CCA and piecewise direct standardization. Copyright © 2014 John Wiley & Sons, Ltd. |
