Detection of mines using hyperspectral analysis

Abstract

This study focuses on the development of computer algorithms that can be used for automatic mine detection using hyperspectral imagery. These algorithms perform a pixel-by-pixel comparison of the scene spectra with the spectrum of a mine. The goal is to assign to every pixel a scale factor which gives the relative probability of finding a mine. Algorithms were tested on simulated data taken from the NPS Middle Ultraviolet Spectrograph (MUSTANG). Three computer methods are tested and relative results were compared. This analysis suggests that the potential exists to use these methods in military applications. The ability to identify features in an image based solely on their spectral signature provides a new dimension to imagery interpretation.