基于D-S证据理论的纹理图像分类方法

在阐述Dempster-Shafer证据理论的基础上,给出了基于Dempster-Shafer证据理论的多源信息融合的方法,并将Dempster-Shafer证据理论的信息融合技术应用于遥感图像纹理的分类.图像灰度均值特征和图像灰度共生矩阵的熵特征作为纹理图像的不同特征被提取,并构成该理论中的证据,利用一定的决策规则,选择融合证据作用下最大的假设.实验结果表明,基于Dempster-Shafer证据理论的多特征融合分类识别图像纹理的新方法是切实有效的和可行的,分类结果要优于仅仅利用单个特征进行分类的结果,能极大地提高图像纹理的识别分类能力.

Classification of texture images based on Dempster-Shafer evidence theory

Based on Dempster-Shafer evidence theory, a method of multi-source information fusion was given which was used in classification of remote sensing texture images. The characteristics of mean of texture image and the ones of the entropy of commensal matrix for texture images were taken as some different characteristics of the texture images, and became the evidence of this theory. It was assumed that the maximum was chosen on the action of fusion evidences in this method by the use of the given decision rules. The experimental results showed that this multi-characteristics fusion classification based on Dempster-Shafer evidence theory is better than other ones by using only one characteristic, improving the classification of texture images.