Date of Award
Master of Science
Department of Engineering Physics
Theodore Luke, PhD
Multispectral detection methods attempt to discriminate targets in a dominant clutter background using multiple images of the same real-world scene taken in different narrow spectral bands in the infrared. Detection is possible due to the empirically observed phenomenon that the radiance of man-made objects, such as a tank or truck, often lies off the main spectral correlation axis of that of natural backgrounds. Radiometric measurements of several vehicles and a tree canopy background taken over three days in June. 1994 were used to examine the factors affecting multispectral detection. Results clearly showed that the processes which provide for higher spectral correlation of natural backgrounds tend to diminish the spectral separation between the background and target. Very high correlations between IR spectral bands for the targets and background were found to exist. The degree of correlation between bands in the LWIR was found to be higher than that for the MWIR for the correlation levels (p) of interest, p ≥ 0.99. The MW-LW combinations were found to almost never produce these high correlations.
DTIC Accession Number
Thomas, Jason E., "Multispectral Detection of Ground Targets in Highly Correlated Backgrounds" (1994). Theses and Dissertations. 6442.