基于信息融合的遥感分类与回归分析研究

Classification and regression analysis of remote sensing images by information fusion

本文提出了组合纹理和光谱信息的遥感图像解译技术,将其运用于陆地水体的二元分类,目的是通过组合两种分析方法的优点,得到更好的解译结果。利用灰度共生矩阵来进行纹理分析,同时,红光和红外光谱分析的使用能够改善陆地和水体的分类边界。纹理分析和光谱特征的融合改进了两类标记的精确度。在此基础上,本文针对多光谱数据的特点提出了灰度级差联合概率矩阵(GLDAP),利用提取的特征对多光谱高度数据进行了回归分析。

A remote sensing interpretation technique combined texture and spectral information, for land/water classification, is given in this paper, in order to get better results taking advantage of both analyses. By using texture based on grey level co-occurrence matrix （GLCM） and red/infrared spectral information, a refined boundary of land/water was obtained, which improved the accuracy of the land type labeling. On this basis, a gray level difference associated possibility matrix （GLDAP）, in view of the characteristics of multi-spectral data, is presented and applied regression analysis to the elevation data of multi-spectral image with the extracted features from GLDAP.