一种新的高光谱遥感蚀变矿物提取算法——MSSS_t相似光谱匹配组合法  

A New Spectral Matching Method for Altered Mineral Identification from Hyperspectral Remote Sensing:MSSS_t

作　　者：韩海辉[1,2,3] 王艺霖 张转[1,2] 任广利 杨敏[1,2,3] HAN Haihui;WANG Yilin;ZHANG Zhuan;REN Guangli;YANG Min(Xi'an Center of China Geological Survey,Xi'an 710054,China;Northuuest China Center for Geoscience Innovation,Xi'an 710054,China;Changfan University,Xi'an 710054,China)

机构地区：[1]中国地质调查局西安地质调查中心,西安710054 [2]西北地质科技创新中心,西安710054 [3]长安大学,西安710054

出　　处：《遥感信息》2019年第5期21-28,共8页Remote Sensing Information

基　　金：国家自然科学基金（41502312）;中国地质调查局项目（DD20160009、DD20190539）

摘　　要：针对高光谱蚀变矿物提取提出的光谱角和光谱信息散度(SIDSAMtan)组合法在复杂地质背景下仍然存在分类精度不足的问题,将波谱特征拟合(SFF)和混合调制匹配滤波(MTMF)加入SID-SAMtan法中,提出MTMFSFF-SIDSAMtan(MSSSt)组合法,可最大化地放大与目标波谱匹配的像元信号。采用相似度增量和相关光谱区分熵指标,定量比较MSSSt、MTMF、SFF、SIDSAMtan、SID、SAM 6种方法对人工模拟的相似光谱的区分能力,发现MSSSt法具有更强的光谱区分能力。在甘肃北山方山口地区,利用新方法从CASI-SASI高光谱影像上提取了绢云母、褐铁矿、绿泥石等9种蚀变矿物,通过地面实测验证对提取结果进行精度评价,发现9种蚀变矿物的提取精度都在90%左右,显示出MSSSt方法行之有效,可作为今后提取高光谱蚀变矿物的方法之一。In recent years,the SIDSAMtan combination method proposed for the extraction of hyperspectral altered minerals still has insufficient classification accuracy under complex geological background.In order to solve these problems,the MTMFSFF-SIDSAMtan(MSSSt)combination method is proposed in this paper.The algorithm adds SFF and MTMF to the known SID-SAMtan model,and the spectral signal of the target pixel is amplified.The similarity increment and spectral distinction entropy index are used to quantitatively compare the similarity of MSSSt,MTMF,SFF,SIDSAMtan,SID and SAM methods.The results show that MSSStmethod has stronger spectral discrimination ability.In addition,nine kinds of alteration minerals,such as sericite,limonite and chlorite,are extracted from CASI-SASI hyperspectral image in Fangshankou area,and the accuracy of nine kinds of altered minerals is generally at about 90% by comparison with ground measured data.The results show that the MSSStmethod is effective,and it can be used as one of the methods to extract altered minerals.

关 键 词：高光谱蚀变矿物 光谱匹配组合算法 相似度增量 相关光谱区分熵 CASI-SASI 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]