Abstract:
OBJECTIVE To explore the color and odor changes of Salvia miltiorrhiza Bge. slices from different origins, and combine modern machine learning technology to achieve rapid differentiation of origins.
METHODS Intelligent sensory technology was used to quantify the color and represent the odor of Salvia miltiorrhiza Bge. slices from different geographical origins. Various data analysis methods including principal component analysis (PCA), discriminant analysis, discriminant factor analysis (DFA), component heat maps, correlation analysis, machine learning and so on, were employed to establish a discrimination function for distinguishing the origin of Salvia miltiorrhiza Bge. slices based on color data.
RESULTS Classification and screening of odor information led to the identification of 10 differential markers: ethanol, carbon disulfide, cyclopentane, 3-methylfuran, propylene glycol, nonane, phenol, 1, 5-octadienone, 1, 8-cineole, and sotolon. It was also found that there was a significant correlation between the color and odor of the slices. Furthermore, based on the concept of data fusion, the study established classification models such as subspace clustering, and compared to single-color discriminant analysis, the classification accuracy was improved to 94.4%.
CONCLUSION The feasibility and superiority of intelligent sensory technology in classifying the geographical origin of TCM is confirmed, providing new methods and insights for quality control of Salvia miltiorrhiza Bge. slices.
下载: