用ASETR图像和地统计学纹理进行岩性分类

运用新获取的ASTER数据可以对岩性进行识别与分类：首先运用地统计学中的变差函数来计算分析几种选定的岩性单元的灰度值空间变化特征；运用ASTER数据的可见光一近红外波段、短波红外(SWIR)波段以及二者的组合进行岩性的分类，分析分类精度的变化。用变差函数作为纹理的计算函数来提取图像纹理，并与原始的光谱数据结合，进行岩性的分类。结果表明，与单纯的光谱分类相比，加入纹理信息可显著改善分类精度；用不同方向的滞后距离提取的图像纹理对图像的分类结果有一定的差异，尤其是对存在明显的各向异性的岩石单元。

LITHOLOGICAL DISCRIMINATION USING ASTER IMAGE AND GEOSTATISTICAL TEXTURE

The use of newly acquired ASTER data for lithological discrimination was evaluated. The variographical analysis was applied to several selected lithological units in order to analyze the spatial variability of these lithological units. Spectral classifications using 3 visible-infrared (VNIR) bands,6 shortwave infrared (SWIR) bands and their combination (9 bands) were performed and evaluated respectively in terms of overall classification accuracy. The texture was extracted from image by using variogram as a texture measure. The texture images,obtained by using different directions of lag distance,specified lag spacing and moving window,were combined with spectral bands for lithological discrimination. The results indicate that the inclusion of variogram texture in spectral classification considerably improves the overall classification accuracy. The effect of the direction of lag distance on variogram texture extraction and subsequential classification was discussed.