Automatic Identification of Herbal Medicines Based on Medicinal Plant Leaf Images Using the Scale Invariant Feature Transform (SIFT) Features

Abstract

Background: A few people prefer to consume medicinal plants compared to modern medicine. This is because modern medicine contains chemicals which over time can have a bad impact on the kidneys, and medicinal plants are also considered cheap treatments. Meanwhile, in our current environment, there are plants that grow and have certain benefits, but some people don't know whether these plants are herbal medicinal plants or not. By utilizing technology, people can find out about herbal medicinal plants based on the leaves by photographing them on an Android smartphone. Method: The method used to extract features from the leaf image is Scale Invariant Feature Transform (SIFT). Aim: This research aims to recognize leaves whose images have been photographed or uploaded. The system will identify herbal medicinal plants using the leaf image of the plant using the Scale Invariant Features Transform (SIFT) method. Result: Feature Extraction and Support Vector Machine (SVM). With this system, it is hoped that users will be able to identify herbal medicinal plants that may grow in the surrounding environment. Based on the description in the background above, the problem formulation in this research is how to identify herbal medicinal plants using leaf images using Android-based SIFT feature extraction. Conclusion: The results of the confusion matrix test explain that this system has an average accuracy of 77%, which means that this system is quite good at identifying leaf images, even though the error rate is quite high at 23%.

Keywords—Medicinal Plant Leafs, SVM, SIFT