Multivariate Calibration of Near Infrared Spectroscopy in the Presence of Light Scattering Effect: A Comparative Study

摘要: When analyzing heterogeneous samples using spectroscopy, the light scattering effect introduces non-linearity into measurements and deteriorates prediction accuracy of conventional linear models. This paper compares performance two categories chemometric methods: pre-processing techniques to remove non-linear calibration directly model non-linearity. A rigorous statistical procedure is adopted ensure reliable comparison. The results suggest that optical path length estimation correction (OPLEC) Gaussian process (GP) regression are most promising among investigated methods. Furthermore, combination models explored with limited success being achieved.