基于线性光谱混合模型的高光谱遥感岩矿分类研究

Classification of Hyperspectral Remote Sensing Rock Based on Linear Spectral Mixture Model

针对高光谱遥感信息冗余多、波段相关性比较强的特点,对预处理的影像进行最小噪声分离(Minimum Noise Fraction,MNF)实现降维和去相关,然后利用纯净像元指数(Pixel Purity Index,PPI)进行端元选择,最终选择斜钠钙石、阳起石、水铝石、斧石、锌蒙脱石5种端元,基于线性光谱混合模型利用IDL语言编程实现Cuprite地区高光谱影像矿物的分类。利用kappa系数对实验分类结果进行精度分析,验证了基于线性光谱混合模型进行高光谱矿物分类的有效性,从而更好地进行矿物识别研究。

In view of the characteristic of high spectrum remote sensing information redundancy and strong band correlation. The Minimum Noise Fraction, MN （MNF transform） is used to reduce the dimension of the preprocessed image. The Pixel Purity Index algorithm for end member selection, final choice of oblique sodium calcium stone, actinolite, aluminum water stone, axinite, zinc montmorillonite five end member. Mineral classification based on linear spectral mixture model using IDL programming. Through the accuracy analysis of the experimental results, it is proved that the method is effective and practical, and it is better to carry out the researchof mineral identification.