结合灰度和基于动态窗口的纹理特征的遥感影像分类

在基于灰度共生矩阵提取遥感影像纹理特征的基础上,针对固定窗口算法的局限性,提出了动态窗口算法;并将不同滑动窗口算法提取的纹理特征与影像灰度组合进行支持向量机(SVM)分类,对分类结果进行定性和定量比较分析。实验结果表明:影像灰度结合动态窗口算法提取的纹理特征进行SVM分类的分类精度优于灰度结合固定窗口算法提取的纹理特征的分类精度。因此,提出的算法较传统的固定窗口算法更具优势,是一种有效纹理信息提取方法。

Classification of Remotely Sensed Imagery according to the Combination of Gray Scale and Texture Features Based on the Dynamic Windows

On the basis of the texture features extracting from the remote sensing images based on gray level co-oc-currence matrix( GLCM) , and aiming at the shortage of conventional algorithms of extracting texture, an improved algorithm of dynamic windows was proposed in this paper. A set of classification experiments were carried out by u-sing the support vector machine( SVM) classifiers with the fusion of gray scale and texture features extracted from different sliding windows. Both qualitative and quantitative approaches were applied to assess the classification re-sults. The experimental results demonstrated that the accuracy of SVM classification combined with texture features based on dynamic windows and gray scale was better than the accuracy of SVM classification combined with gray scale and texture features based on single window. Therefore, the proposed algorithmis superior to traditional algo-rithm and effective for extracting texture information.