Classification of medicinal plant leaf image based on multi-feature extraction

Medicinal plants are the main source of traditional Chinese medicine (TCM), which provides the basic protection of human health. The research and application of medicinal plant classification methodology has important implications in the TCM resource preservation, TCM authentication, and the teaching method of TCM identification. This paper proposes an automatic classification method based on leaf images of medicinal plants to address the limitation of manual classification method in identifying medicinal plants. Our approach will first preprocess the leaf images of medicinal plants; then it will compute the ten shape feature (SF) and five texture characteristics (TF); finally, it will classify the leaves of medicinal plants using support vector machine (SVM) classifier. The classifier has been applied to 12 different medicinal plant leaf images and achieved an average successful recognition rate of 93.3%. The result indicates that it is feasible to automatically classify medicinal plants by using multi-feature extraction of leaf images in combination with SVM. The paper provides a valuable theoretical framework in the research and development of medicinal plant classification system.