Material clustering and band reduction in spectral libraries with unsupervised hierarchical classification methods

In spectroscopy, for material detection and identification applications, spectral libraries are used as reference databases. In spectral libraries, there are spectral reflectance signatures of many materials in electro-optical and optical wavelengths. Sometimes, in spectral libraries, different subclasses of the same materials are included. In material identification applications with hyperspectral images, instead of using directly, spectral libraries can be used after band reduction by using band reduction techniques and the main materials can be estimated by using clustering techniques; in this study, these usage are proposed for a more efficient use of libraries. In this study, for any spectral library, hierarchical clustering methods are proposed for both band reduction and material clustering.