基于最小噪声分量变换的ASTER遥感数据岩性分类

Lithological Classification Based on ASTER Data by Minimum Noise Fraction Transform

遥感岩性分类相较于普通土地利用分类难度更大,其形状或者规模上都不具有特别明显的规律可循,且在边缘的过渡上也存在模糊性,对于常用的监督分类方法在岩性分类中效果都不甚理想。利用ASTER多光谱数据,总结出一套基于最小噪声分量变换的ASTER遥感数据岩性分类技术流程,在ASTER数据最小噪声分量变换的基础上,通过二维散点图几何顶点提取岩性分类样本,利用光谱角技术对西藏日土地区弗野铁矿范围内的岩性进行分类,并通过中值滤波的方法对分类结果进行后处理。结果表明,分类结果与实际岩性有很好的一致性,总体分类精度达到83.33%。

In lithology classification there are not obvious rules from the shape to size.There is no clear dividing line between different lithologies,difficult than land classification by remote sensing data.Lithology classifcation results ars not effective by ordinary supervised classification methods.Our survey uses ASTER multispectral data to summarize a set of technical processes about lithological classification based on the minimum noise fraction transform.According to the geometry vertex of 2D scatter plot chart to extract the lithological classification of samples,spectral angle mapper is used at Fuye iron located in Rutog County of Tibet,in post-processing by median filtering on the classification results.The classification results are in good agreement with the actual situation,and the overall classification accuracy reaches 83.33%.