Hyperspectral Classification Workflows Integrating Dimensionality Expansion for Multispectral Imagery

摘要

Many methods have been developed for target detection and classification in hyperspectral imagery. However, application of these methods to multispectral imagery presents a challenging problem due to a relatively low number of spectral bands available for analysis. Since a multispectral image may contain more targets than spectral bands, there may exist insufficient dimensionality to accommodate the proper classification of all targets in the image scene. To resolve this dilemma, a dimensionality expansion (DE) technique is introduced. These expanded bands ease the problem of insufficient bands in multispectral imagery and can improve and enhance the performance by exploiting band-to-band nonlinear correlation. In this paper, we utilize this DE to apply hyperspectral target detection and classification methods to multispectral imagery. A visual modelling tool is used to build multiple classification workflows using these hyperspectral target detection techniques and to incorporate the DE into the classification process.