辅以纹理特征的面向对象的遥感影像分类方法研究--以陕西省杨陵县为例

The Research of Object-oriented Classification Method in Remote Sensing Image with Texture Analysis——Taking Yangling County of Shanxi Province as an Example

随着遥感图像分辨率的不断提高,图像的纹理特征更加明显,使得对图像进行纹理分析成为可能。以陕西省杨陵县为试验区,将TM多光谱影像和SPOT-5高分辨率全色影像进行融合,先对影像各种纹理特征进行比较,选用对比度这一属性辅助进行面向对象的分割,同时结合地物的光谱信息,对该地区进行监督分类。与单纯基于光谱特征的分类方法进行比较,该方法在一定程度上改善了分类精度,细化了地物类别。

With the resolution enhancing of remote sensing image, the details in the images become clearer, and it makes possible analyzing images with texture feature. Yangling county of Shanxi province is taken as a study area. TM multi-spectral image is merged with high resolution pan SPOT-5 image. Many kinds of texture feature are compared and the contrast feature is selected to help segmenting the image. The spectral information is combined to help classifying this study area. Compared with the classification method purely based on spectral character, this method can improve the classification accuracy and refine object classes in some extent.