Data fusion strategy for rapid prediction of glyceryl trioleate and polysaccharide content in Ganoderma lucidum spore powder based on near-infrared spectroscopy and hyperspectral imaging

From the paper

Abstract

Glyceryl trioleate and polysaccharide are key components in Ganoderma lucidum spore powder (GLSP), which is highly valued for its nutritional value. Traditional methods for assessing these contents are often destructive and time-consuming, while non-destructive techniques like near-infrared spectroscopy (NIRS) and hyperspectral imaging (HSI) offer efficient alternatives, simplifying and accelerating quality evaluations. In this study, NIRS and HSI data of GLSP were collected, and a method based on PLS and low-level data fusion strategy was proposed for predicting GTC and PC. Ten spectral pretreatments, three feature wavelength selection methods, three modeling methods, and two data fusion strategies were applied to build the best predictive models to achieve the rapid and non-destructive quality evaluation of GLSP. GTC’s best model showed optimal performance with R²P of 0.9950, RMSEP of 1.276, and a surprising RPD of 10.01. For PC, its best model had an R²P of 0.9353, RMSEP of 0.1981and a qualified RPD of 2.826. Therefore, results suggested that combining data from NIRS and HSI fusion could rapidly determine the content of active ingredients, overcome the limitations of a single technology, and provide a new perspective for food safety detection.