Nondestructive Testing and Modeling of Chemical Composition of Sporoderm-broken Spores of Ganoderma Lucidum Based on Hyperspectral Imaging Technology and Machine Learning Model

OBJECTIVE To study a non-destructive detection method of chemical composition of sporoderm-broken spores of Ganoderma lucidum based on hyperspectral imaging technology combined with machine learning model.

METHODS  Hyperspectral data of sporoderm-broken spores of Ganoderma lucidum samples from different origins were collected, and backpropagation neural network, extreme learning machine(ELM) and decision tree(DT) models were used to predict the contents of polysaccharides, triterpenoids and ergosterols. In order to improve the prediction accuracy of the model, genetic algorithm(GA) was used for feature wavelength selection, and principal component analysis was used for dimensionality reduction.

RESULTS  The prediction performance based on GA-ELM and GA-DT models was the best, and the prediction determination coefficient(R²) of multiple components was >0.94, showing excellent prediction ability.

CONCLUSION This study verifies the feasibility of hyperspectral imaging combined with machine learning methods for rapid non-destructive detection of the chemical composition of sporoderm-broken spores of Ganoderma lucidum, and provides new ideas and methods for the quality evaluation of sporoderm-broken spores of Ganoderma lucidum.