高光谱遥感岩矿识别方法对比研究

A Comparative Study on the Mineral Identification Methods Using Hyperspectral Remote Sensing Data

高光谱遥感作为全新的遥感技术,在我国目前处于探索研究阶段。本文在高光谱遥感图像处理基础上,对美国Cuprite矿区岩矿开展了基于像元统计和基于波谱特征的分类识别实验研究,提取了八类岩矿分布信息,并应用混淆矩阵评价了不同分类方法、不同岩矿种类的分类精度,研究成果表明:岩矿空间分布越集中,波谱特征越明显,其分类精度越高;分类方法上,基于波谱特征的分类法优于基于传统统计分析的分类法,这为高光谱遥感技术在遥感地质矿物填图的应用提供理论依据。

As a new remote sensing technology, the hypersperctral technique is still in an exploring stage in our country. On the basis of hyperspectral remote sensing image processing,this paper studies the classification and recognition of the Cuprite ore district of the U.S. in view of pixel statistics and spectrum characters, and extracts the information of eight kinds of rocks and ores, and assesses the classification accuracies for different classification meth-ods using confusion matrix. The results show that the more concentrated spatial distribution of rocks and minerals exhibits ,the more obvious spectrum char- acters, and the higher classification accuracy. The classification methods based on the spectral characteristics is better than that based on the traditional statistical analysis. This provides a theoretical basis for mineral mapping using the hyperspectral remote sensing technology.