Rapid identification of Radix Astragali by data fusion of laser-induced breakdown spectroscopy and Raman spectroscopy coupled with deep learning.

The accurate identification of Radix Astragali holds significant scientific importance for evaluating the quality and medicinal efficacy of this herb. In this study, we introduced an efficient methodology, integrating laser induced breakdown spectroscopy (LIBS) and Raman spectroscopy, to identify Radix Astragali samples. Additionally, convolutional neural network (CNN) models were constructed and trained using low-, mid-, and high-level data fusion strategies. The results demonstrated significant improvements in sample classification using all fusion strategies, surpassing the performance achieved when applying LIBS or Raman data individually. Notably, mid-level fusion achieved the highest level of accuracy (93.44 %), with the low- and high-level fusion methods slightly lower at 88.34 % and 90.10 %, respectively. The newly proposed methodology showcased its significance in the rapid and accurate identification of Radix Astragali samples, thereby improving analytical capabilities in Radix Astragali research.