基于峰值权重的岩心高光谱矿化蚀变信息提取

Extraction of mineral alteration information from core hyperspectral images based on weight of absorption peak

针对传统的光谱角制图(spectral angle mapping,SAM)方法在岩心高光谱矿化蚀变信息提取中的局限性,提出了一种基于光谱特征区间吸收峰值权重的高光谱蚀变信息提取方法。首先,在图像反射率转换、噪声去除和端元提取基础上,系统分析了岩心高光谱图像中与铀成矿密切相关的伊利石化、绿泥石化和碳酸盐化等3种铀矿化蚀变的诊断性光谱吸收特征;然后,分别选取同类蚀变光谱曲线之间差异较小的部分作为特征区间,并对特征区间内峰值处的1个波段的反射率设置权重;最后对分别添加权重的参考端元与图像像元进行特征区间内的光谱角计算,实现3种典型铀矿化蚀变信息的提取。该方法突出了吸收峰值和局部光谱信息,能够很好地对同一蚀变进行聚类,区分不同蚀变种类。精度验证和对比结果表明,该方法在权重系数ω=2时,岩心高光谱矿化蚀变信息提取精度可提高20%以上,应用效果显著。

Aimed at overcoming the defects of the spectral angle mapping（ SAM） in extraction of mineral alteration information from core hyperspectral images, the authors proposed an improved method based on a weight in the absorption valley in hyperspectral information identification. Firstly, through necessary reflectivity conversion, noise filtering and selection of end-members in the original image, the authors systematically analyzed the diagnostic absorption features of three typical types of altered minerals closely related to uranium mineralization （ i. e. , illitization, chloritization and carbonatization） in the core hyperspectral image. Then a range called feature range was selected which has little difference between the same kind of altered minerals on their spectral curves. The authors used SAM between end-members and pixels in this feature range after setting up a weight on absorption peaks within a band scope, and finally achieved the recognition and extraction of these three typical uranium mineralization alteration minerals. The method proposed in this paper highlights the important part of spectral information and its absorption peak, and can better cluster the same kind of alterations and distinguish different alteration types. Contrasting and precision test results show that the accuracy can be improved by over twenty percent when ω=2 , and the application effect of extracting information of altered minerals from core hyperspectral image is remarkable.