Applying Tchebichef image moments to quantitative analysis of the components in complex samples based on raw NIR spectra

Abstract

The interferences of irrelevant information, overlapping and shifts of peaks appear mostly in near infrared (NIR) spectroscopy, especially in complex samples, which seriously impede the accurate quantification. In this work, the features of raw NIR spectra represented by Tchebichef image moments (TMs) were employed to partial least square (PLS) modeling. The proposed strategy was applied to quantitative analysis of the components in complex samples based on their raw NIR spectra, and the obtained models were strictly evaluated by their statistical parameters. Our study indicates that the information in raw NIR spectra can be reorganized and represented by TM method owing to its powerful multi-resolution capability and inherent invariance property, which is beneficial to extract the important information of target components. Compared with the PLS and interval partial least square (iPLS) method, the proposed approach could provide accurate and reliable analytical results. Therefore, as an efficient pretreatment method, TMs can be used to improve the analytical precision of PLS based on conventional NIR spectra.