Classification of medicinal plants: An approach using modified LBP with symbolic representation

摘要

In this work, a symbolic approach for classification of plant leaves based on texture features is proposed. Modified Local binary patterns (MLBP) is proposed to extract texture features from plant leaves. Texture of plant leaves belonging to same plant species may vary due to maturity levels, acquisition and environmental conditions. Hence, the concept of clustering is used to choose multiple class representatives and the intra-cluster variations are captured using interval valued type symbolic features. The classification is facilitated using a simple neatest neighbor classifier. Extensive experiments have been carried out on newly created UoM Medicinal Plant Dataset as well as publically available Flavia, Foliage and Swedish plant leaf datasets. Results obtained by proposed methodology are compared with the contemporary methodologies. The Outex dataset is also considered for experiments and the results are promising even on this synthetic dataset. (C) 2015 Elsevier B.V. All rights reserved.