Data analysis strategies for passive multispectral and hyperspectral infrared remote sensors

摘要

The detection of airborne chemicals is a key capability in a variety of environmental monitoring scenarios. For these applications, passive infrared remote sensors collect infrared emissions from natural and manmade sources such as the radiant emission from the earth or emissions from the stacks of a chemical plant. Chemical compounds absorb or emit infrared energy at characteristic wavelengths, and the profile of these absorption or emission signatures can be used to identify a chemical and to estimate the amount present. Passive infrared remote sensors can be implemented in either imaging or non-imaging configurations and can be constructed to acquire infrared emission data in either multispectral or hyperspectral modes. Implementing these measurements successfully requires the construction of rugged, portable instruments and the development of computer processing techniques that allow the automated analysis of the large quantities of data acquired by these sensors. The research presented here describes the development of novel signal processing and pattern recognition methodology for application to multispectral imaging data and to non-imaging data acquired with a hyperspectral instrument. Remote sensing data were collected with these instruments mounted on an aircraft platform. Data acquired at an industrial site are used to demonstrate the characteristics of each sensor and the data analysis methodology.