植被覆盖区卫星高光谱遥感岩性分类  

Lithologic Classification from Hyperspectral Data in Dense Vegetation Cover Area

作　　者：陈圣波[1] 刘彦丽[1] 杨倩[1] 周超[1] 赵靓[1] 

机构地区：[1]吉林大学地球探测科学与技术学院,长春130026

出　　处：《吉林大学学报（地球科学版）》2012年第6期1959-1965,共7页Journal of Jilin University：Earth Science Edition

基　　金：国家自然科学基金项目(40971187);中国地质调查局地质矿产调查评价项目(1212011087112)

摘　　要：植被高覆盖区岩石和土壤在遥感图像上表现为弱信息、小目标,如何利用卫星高光谱遥感提取岩性弱信息是目前遥感地质应用中的最大挑战之一。以黑龙江呼玛地区为例,选择美国EO-1卫星Hyperion高光谱数据。由于植被与下伏岩石-土壤的光谱混合,分别计算研究区含土壤因子和不含土壤因子的植被指数,并对两类不同的植被指数进行主成分分析,以此分离植被和岩石-土壤组分。在含土壤因子植被指数主成分分析的二维组分散点图上,明显区分出背景植被与异常岩石-土壤组分,证实了植被与岩石-土壤组分经主成分分析分离的效果。同时在不添加土壤因子植被指数的分析中,明显区分出植被覆盖信息。通过对实验区典型岩石进行野外光谱测试,然后对光谱进行连续统去除处理,将其作为参考光谱,与分离后的岩石-土壤光谱进行光谱特征拟合(SFF),从而成功地识别出研究区内不同岩石类型,特别是玄武岩、流纹岩、砂砾岩、安山质凝灰岩、大理岩和石英片岩识别效果较好。根据研究区内不同岩石地层单元内岩石组合特征,通过分离后的组分合成图像,成功地实现了岩性分类。与已知地质图叠加,证实通过卫星高光谱数据提取的不同岩石类型颜色边界与地质图岩性界线吻合较好。结果表明:通过植被与岩石-土壤光谱组分分离,结合高光谱遥感的光谱特征拟合,能够识别不同的岩石类型,实现植被覆盖区岩性分类。It is a challenge to identify lithological dense vegetation cover. In this study, Huma area, type by satellite hyperspectral remote sensing under Heilongjiang Province is taken as an example. The Hyperion data from EO - 1 satellite are pre-processed to estimate vegetation indexes. The vegetation cover information, background, rock-soil distribution information and anomaly, are clearly separated on the 2D scattering points diagram generated by employing principal component analysis （PCA） method. By the field spectrum measurements, after the continuum removal of field spectrum from Hyperion image spectrum, main lithological types are identified by spectral feature fit （SFF） method. A technical flow chart is brought forward to classify lithology from hyperspectral data. Applying these methods, the deduced lithologic classification and boundary lines of different geologic units in the study area coincide with the known geology map, proving the ability of the methods in the paper to do lithology classification from hyperspectral data at dense vegetation cover area.

关 键 词：植被覆盖区 高光谱遥感 主成分分析 光谱特征拟合 岩性分类 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]