基于完全谱形特征的成像光谱遥感岩矿识别技术及其应用

DISCRIMINATION TECHNIQUE FOR ROCKS OR ORE DEPOSITS BASED ON THE FEATURE OF FULL SPECTRAL SHAPE USING HYPERSPECTRAL REMOTE SENSING AND ITS APPLICATION

成像光谱以图谱合一、自动生成像元光谱进行直接识别岩矿而极具应用潜力。本文提出并利用基于地物完全谱形特征的成像光谱遥感多波段组合的主成分 (PC,PrincipalComponent Analysis)变换方法提取岩类与蚀变分布 ;相应地利用光谱角识别技术直接提取与矿有关的蚀变矿物 ;最终进行波段组合、波段运算、反差增强与 RGB合成提取出相关岩矿信息进行成矿综合分析。利用岩矿谱形特征进行谱段组合的 PC变换的第一分量 (PC1 )更多地包含岩矿的波谱特征 ;PC2噪声多 ,信息模糊 ,相对信噪比低。利用 PC1提取岩矿信息可靠、准确。这是一种面向目标的定量与半定量相结合的岩矿识别技术 ,在后沟金矿区起到了良好的应用效果。

Hyperspectral remote sensing is used to discriminate and extract widely the information of rocks or ore deposits. This principal composite analysis, which is based on composition band using hyperspectral data according to the full spectral shape feature of different rocks, is very effective to extract the information of different rocks. The first component (PC1) of PC transformation contains more spectral information than others. Suitable RGB composite used PC1 and combined the result of SAM, and then contrast enhancement. It can suggest and show the based rocks, altered rocks and ore deposits. This is a technique which is based on object faced full spectral shape feature to discriminate the information for rocks or ore deposits. Finally, it gains better result at the Hougou gold deposit region than other techniques.