An IoT and Machine Learning Based Intelligent System for the Classification of Therapeutic Plants

The work aim to develop an automatic recognition model using IoT and machine learning (ML) techniques for the classification of therapeutic plants to enrich the traditional medicinal system. Ayurveda, the oldest Indian system of medicine is still practiced today as it widely promotes herbs as medicines for treating different health conditions and presents many advantages, such as low cost, availability in abundance, and minimal side effects. While many countries have accepted conventional medicine as the best alternative to synthetic drugs, the lack of knowledge and unsupported evidence has raised concerns and reduced its usage. Herein, an intelligent system is proposed using Raspberry Pi 3 Model B + (RPi) and the RPi camera to identify real-time images of Indian medicinal herbs and reveal their respective medicinal properties. Four ML models are developed in the work, of which one model is proposed to identify the details of a captured medicinal leaf on the RPi user interface. The proposed model predicts an accuracy (top-1) of 98.98% on a custom leaf dataset of 25 different medicinal species, containing 1500 leaf images, by combining two feature extraction techniques, namely scale invariant feature transform (SIFT) and histogram of oriented gradients. Bag of Visual Words is obtained by applying k-means clustering on SIFT descriptors as a feature selection and assessed using a support vector machine classifier. The suggested model integrated into RPi shows a real-time top-3 accuracy of 99%. The designed system has the advantages of being built solely for medicinal herbs, with reduced camera cost, and even works efficiently in remote areas.