混合像元分解法在植被覆盖区矿化蚀变信息提取中的应用——以江西大浩山金矿区为例

Application of mixed pixel decomposition in mineralization and alteration information extraction in vegetation-covered area: A case study of the Dahaoshan gold deposit in Jiangxi Province

在植被覆盖区,由于高植被、多水系、多阴影的影响,遥感蚀变信息提取一直未能取得良好的矿化蚀变效果。江西大浩山金矿区位于九江坳陷与江南地块九岭隆起的交界部位,区内植被覆盖稠密、水系沟壑纵横,传统的干扰信息掩膜+主成分分析方法难以有效提取相关的矿化蚀变信息。采用混合像元分解技术(SMACC,端元提取技术),通过对像元线性分解可从遥感图像中获取各个地物端元的反射亮度值及丰度信息;识别各地物端元,去除干扰端元(植被、水体等),进而对有用地物端元进行有补偿机制的遥感影像重组;最终采用主成分分析法从重组影像中有效提取矿化蚀变信息。与传统的掩膜方法相比,在植被覆盖率较高的区域,混合像元分解法避免了传统方法在掩膜干扰信息的同时消除矿化蚀变信息的弊端,取得了良好的效果;在植被覆盖率较低的区域,混合像元分解能分离去除少量植被信息,提取更为微弱的矿化蚀变信息。

In the vegetation-covered area,due to high vegetation,multi-water systems and multiple shadows,alteration information extraction of remote sensing image has not achieved good mineralization and alteration results. The Dahaoshan area lies at the junction of Jiujiang Depression and Jiuling Uplift of Jiangnan Block. With dense vegetation and abundant waterways,it is difficult to use traditional″interference information mask ＋ principal component analysis″ method to extract relevant mineralization and alteration information.Thus,this paper used the mixed pixel decomposition technique（ Sequential Maximum Angle Convex Cone,SMACC） to obtain the reflection luminance value and its abundance information of each endmemeber from the remote sensing image by linear decomposition of the pixel. And then it recognized each endmember,wiped off interference endmember（ such as vegetation and water）,then reorganized remote sensing images of useful endmember with compensation mechanism. Lastly,it used principal component analysis to extract the mineralization and alteration information from reorganized images. Compared with the traditional method,it has many advantages： In the densely-vegetated region,the mixed pixel decomposition method avoids the interference information while eliminating the drawbacks of the mineralization and alteration information with good results; In the area of little vegetation,mixed pixel decomposition can separate and remove vegetation information,and extract weaker mineralization and alteration information.