Plant Species Recognition Based on Texture and Geometric Features of Leaf

摘要

This paper proposed a methodology for classification and feature extraction of plant leaves .based on texture and geometric features. In India, a wide variety of medicinal plants are available, manual identification of the leaf is very problematic as it consume more time. Identifying a plant based on leaf features is a challenging task. For plant recognition most significant features such as geometry, shape, texture, color and vein patterns of leaf are utilized. In the first stage of proposed work leaf images are preprocessed to remove noise and to enhance the image that suits for further procedure. Texture features and geometrical features of leaves are computed. Finally, the extracted features are fed to the Multiple Support Vector Machine (MSVM) classifier. The experiments are carried out on a self-built database that contains 1800 images belonging to twenty different plant species. The selected texture and geometric features achieved an accuracy of 96%.