融合元胞自动机和特征加权花卉图像分类方法

A Method of Flower Image Classification Based on Cellular Automaton and Weighted Feature Fusion

图像分割和特征融合是提高花卉图像分类精度的两个主要步骤,但是传统的图像分割方法常常会因花卉图像背景过于复杂而造成分割效果不佳,而且一般的特征融合方法忽略了不同特征对花卉分类贡献的不同。为有效提高花卉图像分类精度,提出一种基于元胞自动机和加权特征融合的花卉图像分类方法。首先,应用元胞自动机在目标和背景之间自然地形成一条明显的分界线,从而将花卉的主体区域从复杂背景中提取出来。其次,对提取的花卉主体区域的颜色特征和局部特征进行加权融合,然后利用SVM实现了花卉图像分类。最后,通过实验验证了该方法对花卉分类的有效性。

Image segmentation and feature fusion are two main steps to improve the precision of flower image classification. But the traditional image segmentation methods often lead to poor effect because of the complicated background of flower images. Furthermore,the general feature fusion method ignores the different contributions of different features on the flower classification. To effectively improve the accuracy of flower image classification,a method of flower image classification based on cellular automation and weighted feature fusion is presented. First,a clear dividing line between the target and background is naturally formed using cellular automaton,and the main area of flowers is extracted from the complex background. Secondly,the color features and local characteristics from the body area are respectively weighted and merged,and the flower images are classified with SVM. Finally,the effectiveness of our method is verified by experiments of flower image classification.