MEDICINAL PLANT IDENTIFICATION: A HYBRID APPROACH COMBINING MANUAL AND DEEP LEARNING FEATURES

Aims: To develop an automated medicinal plant identification system using image processing and deep learning techniques to improve accuracy and reduce reliance on expert knowledge. Study Design: Experimental study involving the classification of medicinal plant species using deep learning and machine learning approaches. Place and Duration of Study: Conducted in an academic research setting over a specific period. Methodology: A dataset of leaf and whole-plant images from six medicinal plant species was created. The approach integrates manually extracted morphological and textural features with deep learning features from a pre-trained ResNet model. These features are classified using a Random Forest classifier to enhance accuracy and robustness. Performance evaluation was conducted based on classification accuracy. Results: The ResNet model achieved 95% accuracy, while the hybrid model combining deep learning and manually extracted features further improved classification performance. Conclusion: The proposed system enhances medicinal plant identification accuracy, preserving traditional knowledge and supporting applications in botanical research, pharmaceutical development, and herbal medicine quality control. Additionally, after identifying the plant, the system provides information about its medicinal benefits, helping users understand its therapeutic applications. Further validation is required to ensure its reliability across diverse datasets.