Distance Metrics and Band Selection in Hyperspectral Processing with Applications to Material Identification and Spectral Libraries

摘要

In this report, we investigate and exploit the properties of distance metrics in hyperspectral processing to achieve superior algorithm performance as well as dimension reduction. Distance metrics are mathematical operators that provide a scalar measure of similarity for two hyperspectral (vector) signals, and they are at the nucleus of many application algorithms. The similarity between two signals, however, can be measured by various means, and different distance metrics offer distinct notions of similarity. Consequently, a thorough understanding of the mathematical and physical properties of distance metrics is crucial to the accurate and efficient processing of hyperspectral data.