中药与临床2025，Vol.16Issue(6)：1-6,6.
基于机器学习的白芷饮片质量等级分类研究
Research on quality grade classification of Angelicae Dahuricae Radix based on machine learning
摘要
Abstract
Objective:This study aims to construct an intelligent classification model for quality grade assessment of Angelicae Dahuricae Radix decoction pieces using machine learning technology,achieving objectification and standardization of quality evaluation,and providing scientific basis for quality control of Angelicae Dahuricae Radix.Methods:A total of 611 batches of Angelicae Dahuricae Radix decoction pieces samples were collected from various regions across the country,and RGB image data were acquired using digital cameras.Traditional image processing techniques were employed to extract morphological,textural,and color features of the samples as model input variables.According to the national standard Commodity Specifications and Grades of Traditional Chinese Medicinal Mate-rials—Angelicae Dahuricae Radix,the samples were divided into three quality grades as model output labels.Quality grade classification models for Angelicae Dahuricae Radix were constructed based on Support Vector Machine(SVM),Convolutional Neural Network(CNN),and Transformer respectively,and the classification performance of each model was compared and evaluated.Results:The study showed that the Transformer model performed optimally with an accuracy of 95.1%,outperforming the CNN model(91.8%)and SVM model(85.2%).SHAP interpretability analysis revealed that color features contributed most significantly to the quality grading of Angelicae Dahuricae Radix,with color mean value being the key discriminative indicator.Ten-fold cross-validation results demonstrated that the model possesses good generalization capability and stability.Conclusion:The intelligent classification model for Angelicae Dahuricae Radix quality grades constructed in this study successfully achieve accurate assessment of Angelicae Dahuricae Radix overcoming the subjective limitations of traditional manual evaluation.It provides objective and efficient technical means for quality control of traditional Chinese medicine,and has important application value for promoting standardization and intelligent detection of traditional Chinese med-icine quality.关键词
白芷/特征提取/质量等级分类/机器学习/图像处理/质量控制Key words
Angelicae Dahuricae Radix/feature extraction/quality grade classification/machine learning/image processing/quality control分类
医药卫生引用本文复制引用
聂蔚,袁宏,任超翔,陈翠平,黄旭龙,裴瑾..基于机器学习的白芷饮片质量等级分类研究[J].中药与临床,2025,16(6):1-6,6.基金项目
四川省科技计划重点研发项目(2023YFS0338) （2023YFS0338）
