Rapid identification of Astragalus membranaceus processing with rice water based on intelligent color recognition and multi-source information fusion technology

Abstract

Objective

This study seeks to optimize the processing parameters for Astragalus membranaceus with rice water (AM-RW), establish quality evaluation standards, and develop a rapid multilayer perceptron (MLP) model for classification. This model facilitates accurate identification of AM-RW at various processing stages, providing a scientific reference for the quality assessment of traditional Chinese medicine products.

Methods

Optimization of AM-RW was achieved using a single-factor test and Box-Behnken design response surface methodology to determine the optimal process parameters. The Watershed Algorithm was applied to segment images of AM tablets, and the numpy and pandas libraries were used to collect color data from these tablets. The study also explored the correlation between R, G, B, and L color values and calycosin-7-glucoside content. A rapid classification model based on MLP was developed, utilizing color values, hardness values, and calycosin-7-glucoside content of AM-RW with various processing degrees.

Results

The study identified the optimal parameters for AM-RW as 20 mL of rice water, a frying temperature of 180 °C, and a frying time of 7 min. The average color values for the best-processed products fell within the normal distribution range: R value (94.83 ± 8.57), G value (96.1 ± 19.37), B value (36.84 ± 5.93), and L value (89.55 ± 12.24). The rapid identification model using MLP demonstrated high accuracy and reliability, achieving an accuracy rate of 94% in the classification process.

Conclusions

The response surface method effectively optimizes the precise processing parameters of AM-RW. Furthermore, the MLP-based model can accurately classify AM-RW with varying degrees of processing, providing a valuable reference for the expedited identification of processing quality in traditional Chinese medicine products.