Non-Destructive Testing Model of Flavonoids / Polysaccharides in Anoectochilus roxburghii By Hyperspectral Imaging Technology

A hyperspectral model for the testing of flavonoids and polysaccharides in Anoectochilus roxburghii (Wall.) Lindl. was developed using Hyperspectral imaging technology and spectral data analysis technology. In this study, the A. roxburghii grown under 7 kinds of LED light quality were used as samples, its spectral data and flavonoid and polysaccharide content data are fitting by multiple linear regressions (MLR), partial least-squares regression (PLSR), and principal component regression (PCR) algorithms, and then these data were used to establish a testing model for the content of flavonoids and polysaccharides of A. roxburghii, as well as to optimize and verify the modeling method. The optimal result of the modeling method in this paper shows that the MLR model based on characteristic wavelengths is the dominant model that better than PLSR and PCR model. Before building the MLR model, the preliminary pre-selection of hyperspectral data preprocessing was performed, and different preprocessing methods were tried to optimize the model. The results show that the testing models for flavonoids and polysaccharides are similar-the effect of using black-and-white to correct hyperspectral data for MLR modeling was already good enough, and there was no more prominent advantage of adding more preprocessing. In this study we can know it is feasible to fit the hyperspectral data to chemical composition data of the samples and then establish a non-destructive testing model.