Ayurvedic Plant Leaf Detection Using HOG Feature Descriptor and SVM Classifier

This study addresses the vital task of identifying plant species, focusing specifically on medicinal herbs, a critical endeavor for biodiversity conservation and advancements in herbal medicine. Utilizing a Mendeley dataset, the research employs a meticulous methodology to enhance classification accuracy. Initial preprocessing involves grayscale conversion and Gaussian filtering, followed by feature extraction using the Histogram of Oriented Gradients (HOG) descriptor. To address bias, Minmax normalization is applied, and feature selection and dimensionality reduction employ Neighborhood Component Analysis (NCA) and Principal Component Analysis (PCA). The study seamlessly integrates image processing, feature extraction, normalization, and classification techniques, employing a Support Vector Machine (SVM) classifier for efficient plant leaf recognition. This holistic framework, devoid of deep learning, demonstrates practical implications, including the potential development of a mobile application bridging scientific research and public awareness. Overall, the research offers a comprehensive approach, leveraging image processing and machine learning to advance leaf classification accuracy in the absence of deep learning techniques.