Automatic Identification of Botanical Samples of leaves using Computer Vision

Abstract

Leaf can be one of the many different parameters on the basis of which a plant can be uniquely identified. Many plants types are on the verge of extinction and can be taken care of, if identified correctly. The proposed method discusses an automated image processing system for leaf classification. The leaf pixels from the image are segmented and termed as region of interest (ROI). A set of geometrical, textural and statistical features is extracted for each input sample and analyzed using a multi class SVM classifier. The proposed system has achieved an accuracy of 97% with a sensitivity of 98.32%. The results are encouraging for a dataset consisting of 10 different leaf classes and can be used for development of some real time application.