Representation and Classification of Medicinal Plant Leaves: A Symbolic Approach

摘要

In this paper, a method of representing medicinal plant leaves shape in terms of tri-interval valued type symbolic features is proposed. The proposed method exploits the axis of least inertia of a leaf shape to extract invariant features for its characterization. Concept of clustering is used to select class representatives for each medicinal plant species. Multiple class representatives are selected to incorporate intra-class variations in each species to facilitate the task of classification. To study the efficacy of the proposed representation and classification technique, experiments are conducted on Flavia data set. The results obtained are more encouraging and a comparative study with the results of other contemporary methods on the same data set is presented.