基于高光谱成像的丹参饮片多指标成分原位快速定量方法OA

Objective This study aimed to develop an in-situ and rapid quantitative analytical method for five key components of Salvia miltiorrhiza slices using hyperspectral imaging(HSI),and to systematically evaluate the impact of different spectral band selections and feature wavelength screening strategies on the performance of partial least squares regression(PLSR)models.Methods Hyperspectral images in the visible-near-infrared and near-infrared bands were acquired from 115 batches of Salvia miltiorrhiza slices.The contents of five key components—salvianolic acid B,tanshinone I,cryptotanshinone,tanshinone ⅡA,and total tanshinones—were quantified.Spectral data from the VNIR,NIR,and fused full-band ranges were extracted.These spectra were preprocessed using six methods,including multiplicative scatter correction and standard normal variate transformation.Feature wavelengths were then selected by competitive adaptive reweighted sampling,successive projections algorithm,and uninformative variable elimination.Finally,partial least squares regression models were developed and evaluated based on the correlation coefficient,root mean square error of the validation set,and residual predictive deviation.Results The partial least squares regression model,built using full-band spectra and refined by competitive adaptive reweighted sampling,achieved optimal predictive performance.For all five analytes,the validation correlation coefficients(Rv)ranged from 0.92 to 0.96,and the residual predictive deviation values exceeded 3.5,demonstrating enhanced model robustness and reliable predictive capability.Conclusion The integration of full-band hyperspectral data with CARS feature selection enables rapid,nondestructive,and accurate quantification of five key components in Salvia miltiorrhiza slices.This approach provides a scientific basis and technical support for intelligent quality control and rapid on-site testing of traditional Chinese medicinal slices.