HYBRID CNN–DECISION TREE FRAMEWORK FOR MEDICINAL LEAF RECOGNITION

Abstract

Accurate and automated identification of medicinal plant species is crucial for various applications, including pharmaceutical research, agricultural practices, and the preservation of ethnobotanical knowledge. Traditional manual identification methods are often time-consuming, prone to human error, and require specialized taxonomic expertise, limiting their scalability and accessibility. This study addresses these challenges by developing and comparatively evaluating two machine learning approaches Convolutional Neural Networks (CNNs) and Decision Trees (DT) for the recognition of medicinal plant leaves. Our proposed system leverages deep learning techniques to analyze visual features of leaf images, enabling efficient and precise classification. The research involved a comprehensive process of dataset collection, meticulous preprocessing, model training, and rigorous performance evaluation using standard metrics such as accuracy, precision, recall, and F1-score. We incorporated both automated feature extraction by CNNs and handcrafted features, including Gray Level Co-occurrence Matrix (GLCM) for Decision Trees, to provide a robust comparative analysis. The CNN model demonstrated superior performance, achieving an accuracy of 97.2% compared to the Decision Tree's 85.6%, alongside higher precision, recall, and F1-scores. This highlights the CNN's robustness and its capability to learn intricate hierarchical features from complex visual data. The findings suggest that CNNs are highly suitable for practical deployment in real-world scenarios, contributing significantly to improved accessibility, preservation, and application of traditional medicinal knowledge through advanced automated identification systems.