傅里叶谱纹理和光谱信息结合的高分辨率遥感影像地表覆盖分类

发布时间：2018-10-12 19:27

【摘要】：以高空间分辨率遥感影像为研究对象,将纹理特征与影像的光谱特征结合起来,用于地表覆盖类型分类。设计了一种基于傅里叶谱纹理的分类策略,对主成分分析后的第1、2主分量特征影像,利用径向谱(r-spectrum)提取纹理特征,并将纹理与光谱特征结合起来,构建了不同的分类特征用于支持向量机分类模型。以Salinas数据集和QuickBird影像为例,验证该算法。结果表明,纹理与光谱信息的结合可以明显提高高分辨率遥感影像的分类精度;由傅里叶径向谱提取的纹理特征可以很好的应用到高分辨率遥感影像的分类问题中,分类精度高于基于傅里叶总能量谱和灰度共生矩阵的分类精度;利用该算法对PCA变换后的第1和第2分量提取的纹理特征具有一定的互补性,并且结合多特征图像的纹理特征提取优于单特征图像的纹理特征提取。
[Abstract]:The high spatial resolution remote sensing image is taken as the research object. The texture feature is combined with the spectral feature of the image to classify the land cover type. In this paper, a classification strategy based on Fourier spectrum texture is designed. For the first principal component feature image after principal component analysis, radial spectrum (r-spectrum) is used to extract texture feature, and the texture is combined with spectral feature. Different classification features are constructed to support vector machine classification model. Taking Salinas dataset and QuickBird image as examples, the algorithm is verified. The results show that the combination of texture and spectral information can obviously improve the classification accuracy of high-resolution remote sensing images, and the texture features extracted from Fourier radial spectrum can be applied to the classification of high-resolution remote sensing images. The classification accuracy is higher than that based on Fourier total energy spectrum and gray level co-occurrence matrix, and the algorithm is complementary to the texture features extracted from the first and second components of PCA transform. The texture feature extraction of multi-feature image is better than that of single feature image.
【作者单位】：华中农业大学园艺林学学院;武汉大学遥感信息工程学院;
【基金】：国家863计划(2012AA12A304,2013AA102401) 中央高校基本科研业务费专项基金(2014QC018) 地理国情监测国家测绘地理信息局重点实验室开放研究基金(2013NGC05)~~
【分类号】：P237

【相似文献】