Using Near-Infrared Hyperspectral Imaging Combined with Machine Learning to Predict the Components and the Origin of Radix Paeoniae Rubra

Abstract

Abstract: The efficacy and safety of drugs are closely related to the geographical origin and quality of the raw materials. This study focuses on using near-infrared hyperspectral imaging (NIR-HSI) combined with machine learning algorithms to construct 9 content prediction models and 9 origin identification models to predict the component and origin of Radix Paeoniae Rubra (RPR). These models can be deemed as quick, non-destructive, and accurate to predict the content and origin of indicator components. This study preprocessed spectral data using three methods: multiple scattering correction (MSC), Savitzky-Golay (S-G), and standard normal variate (SNV). Principal component regression (PCR), partial least squares regression (PLSR), and ridge regression (RR) were used to establish a content prediction model for paeoniflorin. Classification models for the identification of the origin of RPR were developed using three algorithms: support vector machine (SVM), K-nearest neighbor (KNN), and random forest (RF). SNV-RR and SNV-SVM were respectively the most accurate quantitative model and recognition model. The determination coefficient of the SNV-RR model prediction set is 0.8943, and the accuracy of the SNV-SVM model is 0.9790. To simplify the predictive model and enhance computational efficiency, feature wavelengths can be selected for model construction. These feature wavelengths significantly simplify the model structure while maintaining the performance of the full-spectrum model and reducing computational costs. This study demonstrates the feasibility of combining NIR-HSI with machine learning in the quality analysis of RPR. It provides a theoretical basis for the promotion and application of hyperspectral imaging technology in the fields of food and medicine.

Keywords: Keywords: NIR-HSI, machine learning, Radix Paeoniae Rubra, origin identification, content prediction