Visualization Techniques for Multi- and Hyperspectral Imagery Exploitation

Hyperspectral and multi-spectral sensors are assuming a greater role in all aspects of military surveillance and intelligence. While many excellent packages exist for analyzing this data, there has been minimal progress made towards the development of techniques that focus on supporting/aiding the hyperspectral imagery analyst in procedures for semi-automatic extraction of targets. Imagery analyst often rely on scene context to aid in identification of targets, and the identification of scene elements such as road surface, vegetation, building materials, etc. is not a part of target detection software. There is a major divide between scene classification techniques that are developed for terrain classification and target detection algorithms that are often developed for real-time sensors. We propose to develop a software environment that will provide both advanced target detection capability as well as the ability to use scene classification to determine the context of the target.