Translator Disclaimer
13 November 1996 Use of hyperspectral imagery for broad-area detection of small targets
Author Affiliations +
There is a widespread perception that hyperspectral imagery may be useful for military broad area tactical surveillance applications in the not-too-distant future, and the government is planning to demonstrate this technology from space within the next three years. In order to support sensor design for such a demonstration, we have developed an algorithm performance simulation that applies linear unmixing techniques to the problem of quasi-real-time ground clutter suppression in hyperspectral images. Our object is to enable top level sensor design tradeoffs to be evaluated and to derive first order estimates of sensor requirements that could potentially enable fully automated timely detection of time-critical tactical military targets over broad areas against a wide variety of terrain types. We describe a simple algorithm for target detection that is single-pass, completely automated, and requires little or no training to detect targets. Monte Carlo simulations on AVIRIS images have been used to measure the performance of this algorithm under realistic conditions and to derive probability of detection and probability of false alarm versus signal to noise ratio in a pair of test images. Although the parameter space is quite large and the exploratory work is still in progress, early results give valuable insight into design requirements for spaceborne hyperspectral sensors to support broad area tactical surveillance applications.
© (1996) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Walter F. Kailey and Luanna R. Illing "Use of hyperspectral imagery for broad-area detection of small targets", Proc. SPIE 2819, Imaging Spectrometry II, (13 November 1996);

Back to Top