Scholarlink

<p>Alismatis Rhizoma (AR) is widely utilized as a natural medicine across many Asian countries. However, in China, due to its complex origins, AR quality varies, which can affect clinical efficacy. Therefore, there is a need for a method that is both fast and objective to determine the source of AR. In this study, a total of 400 samples of two species and four geographic origins from AR were imaged and processed. From these images, 17 features were extracted, including three shape (S), two color (C), and 12 texture features (T), resulting in a total of 6800 data points. Four commonly used classification models Random Forest (RF), Extreme Learning Machine (ELM), Back Propagation (BP) neural network, and Support Vector Machines (SVM) were tested to find the optimal combination of AR fusion features and classification models. The S + T-RF combinations achieved the best results, with 99.17% accuracy in two species identification and 96.67% accuracy in four geographic origin identification on test sets. These results suggest that image processing combined with the RF model can quickly and effectively identify the complex origins of AR and can provide a reference for the origins identification of other natural medicines.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Intelligent identification method of origin for Alismatis Rhizoma based on image and machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年4月23日

查看全文
  • Wenqi Zhao,
  • Zongyi Zhao,
  • Wen Zheng,
  • Zimin Wang,
  • Gaoting Yang,
  • Zhiqiong Lan,
  • Xiaoli Pan,
  • Min Li

摘要

Alismatis Rhizoma (AR) is widely utilized as a natural medicine across many Asian countries. However, in China, due to its complex origins, AR quality varies, which can affect clinical efficacy. Therefore, there is a need for a method that is both fast and objective to determine the source of AR. In this study, a total of 400 samples of two species and four geographic origins from AR were imaged and processed. From these images, 17 features were extracted, including three shape (S), two color (C), and 12 texture features (T), resulting in a total of 6800 data points. Four commonly used classification models Random Forest (RF), Extreme Learning Machine (ELM), Back Propagation (BP) neural network, and Support Vector Machines (SVM) were tested to find the optimal combination of AR fusion features and classification models. The S + T-RF combinations achieved the best results, with 99.17% accuracy in two species identification and 96.67% accuracy in four geographic origin identification on test sets. These results suggest that image processing combined with the RF model can quickly and effectively identify the complex origins of AR and can provide a reference for the origins identification of other natural medicines.