Table of Contents
Understanding Hyperspectral Imagery
Spectral Profile Tool
View My Guestbook
Sign My Guestbook
©Rob Whittington CMST 386 Section 6380 Spring Semester 2002 University of Maryland University College 10 July 2002
In today's world of surgically precise military strikes, knowing what to target and what to protect is critical to a military campaign's success not only on the battlefield but also on the world's political stage. For this reason, automatic target recognition through spectral data analysis is not only one of the most promising military applications of hyperspectral imagery, but probably one of the most important too.
According to the Computer Vision Laboratory at University of California Irvine, automatic target recognition systems that use only spatial information are severely limited in their ability to detect targets with low contrast or in environments containing camouflage or deception. The Computer Vision Laboratory went on to say, "that the development of hyperspectral sensors promises to provide imagery that can be exploited to improve the performance of target recognition systems in these circumstances" (9). Key to any target recognition system is what is known as a spectral profile tool.
A spectral profile tool (Click the thumbnail image at left to see an example of what one looks like!) provides a graphical depiction of an object's returns at each of the recorded hyperspectral bands. The Spectral Profile tool found within ERDAS Imagine; a powerful, low-cost, stand-alone image mapping and visualization product; "provides a spectral profile at a specific geographic location, drilling down through all the bands in an image and producing a scaled graph of reflectance versus wavelength. Erdas' Spectral Profile tool is most often used for analyzing hyperspectral imagery by comparing the spectral profile of a pixel with a library of known spectra to identify a material at that location" (10).
Since all objects create unique hyperspectral returns, you may liken a spectral profile tool's graphical depiction to a fingerprint-a fingerprint allowing rapid, precise identification of targets. Armed with a catalog of spectral fingerprints, imagery exploitation computers will automatically recognize targets by analyzing their spectral returns and comparing them to the catalog and allow near-real-time identification of friends and foes alike.