Global Scenario of Medicinal Plants

Abstract

Agro-industries have been pushed to the brink by the rapid increase in demand for medicinal plants with pharmaceutical importance and the many ayurveda or herbal remedials. However, increased instances of plant diseases have capped aggregate growth, reducing output volume and quality. In this research, we provide the first hybrid deep-spatial temporal textural feature learning model for medicinal plant disease detection (HDST-MPD), which is powered by evolutionary computing. The HDST-MPD model originally used firefly heuristic driven fuzzy C-means clustering to obtain ROI-specific RGB areas, which helped to reduce the likelihood of a class-imbalance issue occurring. Then, it used the AlexNet transferrable network and the gray-level co-occurrence matrix (GLCM) to make the most of the deep spatiotemporal textural data. In this case, high-dimensional features were generated using the AlexNet deep model, and the inclusion of numerous GLCM features aided in leveraging the distribution of textural characteristics. Each sample medical picture was labeled as either "normal" or "diseased" using a composite vector trained on a random forest ensemble using these deep-spatial, temporal, textural feature (deep-STTF) characteristics. In-depth performance evaluation showed that the suggested model is very effective at real-time illness detection and classification in medicinal plants, with an accuracy of 98.97%, precision of 99.42%, recall of 98.89%, F-measure of 99.15%, and an equal error rate of 1.03%.
Keywords: AlexNet , Gray-level co-occurrence matrix ,Heuristic driven segmentation, Hybrid deepTTF feature , Medicinal plant disease detection