基于灰度共生矩阵的图像纹理特征地物分类应用

Application of GLCM-Based Texture Features to Remote Sensing Image Classification

针对传统遥感影像分类方法的分类精度不高,在分析图像的光谱信息的基础上,对基于灰度共生矩阵的纹理特征在地物分类中的应用进行了研究。本研究利用原始图像进行主成分分析后的前两个主成分,经过编程运算,提取了基于灰度共生矩阵方法的不同测度的纹理特征,将提取的纹理特征作为新的波段,与原始波段进行组合,再对组合图像进行监督分类,探索了利用纹理特征进行地物分类的可行性,并将分类结果与最大似然法分类结果进行定性和定量比较分析。结果表明,综合了纹理特征和光谱特征的地物分类方法,能够有效地提高地物分类精度,证明了基于纹理特征的遥感影像分类的有效性。

In order to solve the problem of low-accuracy in the conventional classification of remote sensing image classification, a new method based on gray level co-occurrence matrix（GLCM） texture features is presented and utilized. After the principal component analysis, the first two principal com- ponents were selected to extract the texture features of different measurements based on gray level co-occurrence matrix. As a new band, the extracted texture features with the original bands were combined by supervised classification methods to classify the images. The classification result was compared with the maximmn likelihood classification qualitatively and quantitatively. The research results show that a combination of texture features with spectral charac- teristics can enhance the accuracy of ground object classification, and prove the feasibility and effectiveness of the GLCM-based classification method presented in this paper.