A new scattering correction method of different spectroscopic analysis for assessing complex mixtures.

Physical properties such as particle size distribution and compactness have significant confounding effects on the spectral signals of complex mixtures, which multivariate linear calibration methods such as partial least-squares (PLS) cannot effectively model or correct. Therefore, these effects significantly deteriorate calibration models' predictive abilities for spectral quantitative analysis of complex mixtures. Here, new scattering correction methods were proposed to estimate the additive and multiplicative parameters considering light scattering effects in each spectrum and hence mitigate the detrimental influence of additive and multiplicative effects on quantitative spectroscopic analysis of complex mixtures. Three different correction methods were proposed to estimate the addition coefficient based on two different underlying assumptions, namely, whether this coefficient is related to the wavelength. After addition coefficient elimination, the multiplicative parameter can be eliminated by a simple but very efficient spectral ratio method. Furthermore, linear models are built with key variables, and the predictive performance of these models is verified using the root-mean-square error of prediction datasets. The proposed methods were tested on one apple data set and two publicly available benchmark datasets (i.e., near-infrared spectral data of meat and powder mixture samples) and compared with some existing correction methods. The results showed that (1) additive effects of different types of samples can be eliminated by different methods and (2) these methods can appreciable improve quantitative spectroscopic analysis of complex mixture samples. This study indicates that accurate quantitative spectroscopic analysis of complex mixtures can be achieved through the combination of additive effect elimination and the spectral ratio method.