联合纹理和光谱特征的高光谱图像分类方法

Texture-Spectrum Joint Classification Method for Hyperspectral Remote Sensing Image

高光谱图像分类是当前遥感信息处理的热点问题.传统高光谱遥感图像分类方法只利用图像的光谱特征,没有考虑高光谱遥感图像各像素点邻域的空间特征.文中提出了一种联合纹理特征与光谱特征的高光谱图像分类方法.首先,使用灰度共生矩阵提取了高光谱遥感图像每一像素点邻域的贡献较大的六个纹理特征,再联合各像素点的光谱特征,形成纹理-光谱特征.最后,基于支持向量机和极端随机树算法对公开的高光谱遥感图像数据集Indian Pines和Pavia University scene进行分类实验,结果表明该方法相比传统方法取得更高的分类性能.

In remote sensing information processing area, hyperspectral image classification is a hot topic. The hyperspectral remote sensing image classification method uses only the spectral features of the image,without considering the spatial features of the neighborhood of each pixel in hyperspectral remote sensing images.In this paper, a texture-spectral joint classification method for hyperspectral remote sensing image was proposed. First, it utilized the gray level co-occurrence matrix to extract six large texture features from each pixel neighborhood of hyperspectral remote sensing images, and then combined the spectral features of each pixel to form texture-spectral joint features. Last, support vector machine and extremely randomized trees classification algorithm of hyperspectral remote sensing image were used in the experiment on the public data set Indian Pines and Pavia University based on scene, the results showed that compared with the traditional method the proposed method had higher classification performance.