Scholarlink

<p>Medicinal plants are very important sources of bioactive compounds that are used in traditional and modern medicinal preparations. To determine the safety, authenticity and effectiveness of natural remedies, correct identification of medicinal plant species is necessary. This paper presents HerbFusionNet, a state-of-the-art Convolutional Neural Network (CNN) architecture, which is developed to handle automated leaf classification of medicinal leaves. The proposed model incorporates the Holistically Nested Edge Detection (HED) to identify the intricate leaf morphology and venation configurations at a high level of accuracy. The VGG-16 architecture is applied to extract the deep features and optimized based on the Chi-square-based feature selection to filter the most discriminative features. Moreover, to optimize the parameters of the CNN, a hybridization approach of the Gray Wolf Optimization (GWO) and Whale Optimization Algorithm (WOA) is utilized in order to increase the robustness and convergence of the model. On a real-world dataset of 26 medicinal plant species, experimental analysis exhibits consistent and high-performance, with an accuracy rate, precision rate, recall rate, sensitivity rate and F1-score of 94, 94, 94, 94 and 94, respectively. These findings validate that HerbFusionNet can provide a balanced, bias-free operation and is a substantial improvement over traditional classifiers, including KNN, SVM, and Gradient Boosting, and has the potential to be used in the real-world of medicinal plant identification and digital herbarium.</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

Deep Learning-Based Medicinal Plant Identification Using Optimized CNN Architectures

  • 期刊论文

  • 发表日期：

    2025年11月15日

查看全文
  • S. Rajani,
  • M. N. Veena

摘要

Medicinal plants are very important sources of bioactive compounds that are used in traditional and modern medicinal preparations. To determine the safety, authenticity and effectiveness of natural remedies, correct identification of medicinal plant species is necessary. This paper presents HerbFusionNet, a state-of-the-art Convolutional Neural Network (CNN) architecture, which is developed to handle automated leaf classification of medicinal leaves. The proposed model incorporates the Holistically Nested Edge Detection (HED) to identify the intricate leaf morphology and venation configurations at a high level of accuracy. The VGG-16 architecture is applied to extract the deep features and optimized based on the Chi-square-based feature selection to filter the most discriminative features. Moreover, to optimize the parameters of the CNN, a hybridization approach of the Gray Wolf Optimization (GWO) and Whale Optimization Algorithm (WOA) is utilized in order to increase the robustness and convergence of the model. On a real-world dataset of 26 medicinal plant species, experimental analysis exhibits consistent and high-performance, with an accuracy rate, precision rate, recall rate, sensitivity rate and F1-score of 94, 94, 94, 94 and 94, respectively. These findings validate that HerbFusionNet can provide a balanced, bias-free operation and is a substantial improvement over traditional classifiers, including KNN, SVM, and Gradient Boosting, and has the potential to be used in the real-world of medicinal plant identification and digital herbarium.