21 May 2015 Virtual dimensionality analysis for hyperspectral imagery
Author Affiliations +
Abstract
Virtual dimensionality (VD) has been widely used to estimate number of endmembers in the past. Unfortunately, the original idea of VD was developed to specify the number of spectrally distinct signatures in hyperspectral data where there is no provided specific definition of what “spectrally distinct signatures” are. As a result, many techniques developed to estimate VD have produced various values for VD. This paper addresses this issue by develops a target specified VD (TSVD) theory where the value of VD is completely determined by targets of interest. In particular, the VD techniques can be categorized according to targets characterized by eigenvalues/eigenvectors and real target signal sources which are used for a binary composite hypothesis testing problem. For the latter case the Automatic Target Generation Process (ATGP) is particularly used to generate real target signal sources to replace eigenvalues/eigenvectors as signal sources to be used for the binary hypothesis testing problem. In order to find probability distributions under each hypothesis the extreme theory used by Maximum Orthogonal Complement Algorithm (MOCA) is used for their derivations. As a result, VD can be estimated by two types of signals sources, eigenvalues/eigenvectors along with two types of detectors, maximum likelihood detector and Neyman-Pearson detector.
© (2015) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Chein-I Chang, Li-Chien Lee, Drew Paylor, "Virtual dimensionality analysis for hyperspectral imagery", Proc. SPIE 9501, Satellite Data Compression, Communications, and Processing XI, 95010R (21 May 2015); doi: 10.1117/12.2176772; https://doi.org/10.1117/12.2176772
PROCEEDINGS
11 PAGES


SHARE
Back to Top