A method for regional feature extraction of flower images

Abstract

The classification of flowering plants is a significant part of plant taxonomy. Automatic recognition of flowers using computer technology is of great social benefits. We analyze the features and disadvantages of existing methods, and propose a regional feature extraction method based on the shape characteristics of flowers. Since most flowers bloom radially and provide centrosymmetric features for their images, we therefore divide each flower image into several feature-ring regions and define their correspond features according to the information contained in them. Experimental results show that a feature system that can effectively distinguish flowers could be established on the basis of the features of flower images extracted through the proposed method, with which a high classification and recognition fidelity of flowering plants can be reached.