Title :
Generalized Chromodynamic Detection
Author :
Stocker, Alan ; Villeneuve, Pierre
Author_Institution :
Space Comput. Corp., Los Angeles, CA
Abstract :
A Generalized Likelihood Ratio (GLR) detector is formulated for a known spectral contrast signal that may be present with unknown amplitudes in a pair of co-registered hyperspectral image observations. Under multivariate Gaussian hypotheses, the joint likelihood function is a quadratic expression containing the signal coefficients for the two spectral observations. Maximum-likelihood coefficient estimates, obtained by solving a set of linear equations, define a multi-look GLR decision statistic for detecting the assumed target signal. The performance of the new multi-look detector is compared to several existing methods using a real LWIR hyperspectral image pair with injected target signals simulating gas effluent plumes.
Keywords :
aerosols; atmospheric techniques; geophysical signal processing; image registration; maximum likelihood estimation; remote sensing; signal detection; spectrochemical analysis; GLR detector; Generalized Likelihood Ratio detector; coregistered hyperspectral image pair; generalized chromodynamic detection; joint likelihood function; maximum likelihood coefficient estimates; multilook GLR decision statistic; multilook detector; multivariate Gaussian hypotheses; signal coefficients; spectral contrast signal; target signal detection; Concatenated codes; Covariance matrix; Detectors; Effluents; Hyperspectral imaging; Matched filters; Optical noise; Pixel; Statistics; Testing;
Conference_Titel :
Geoscience and Remote Sensing Symposium, 2008. IGARSS 2008. IEEE International
Conference_Location :
Boston, MA
Print_ISBN :
978-1-4244-2807-6
Electronic_ISBN :
978-1-4244-2808-3
DOI :
10.1109/IGARSS.2008.4779066