Study On The Flower Identification Technology With Digital Images

Plant classification is groundwork in botanical studies, and in agro-forestry production and management. Plant taxonomy is a fundamental research field whose classification criteria rely heavily on the appearances of plants such as leaves, flowers, branches, bark, and fruits. The classification of flowering plants is a significant part of plant taxonomy. Automatic recognition of flowers with high fidelity using computer technology is of great social benefits.This paper explores the method for flowering plant classification and recognition with the digital images of some common ornamental flowers. Based on previous studies, this paper presented the features of color, texture and shape of the flower image based on growth characteristics of flowers, and using a hierarchical SVM classifier to classify images of flowers. This paper proposes a regional feature extraction method and Gary Level Co—Occurrence Matrix in polar coordinate system for the centrosymmetric image like flower image, which is significant for improving the system's recognition accuracy. In addition, the proposed hierarchical SVM classifier effectively reduced the sensitivity of classifier to number of the types of samples, which solved the problem that the recognition accuracy is low when SVM classifier is used to classify samples with large number of types. This paper built a digital image-based classification system of flowers, and used images of fifty kinds of flowers to test the system, got the finally classification rate of 90.8%. The experiment result showed that the achieved flower classify system has high recognition accuracy and stability.