Signal and Image Processing Techniques for Environmental and Clinical Applications of Infrared Spectroscopy

by Wabomba, Mukire John

Abstract (Summary)
The development of automated, real time, and robust measurement techniques is the focus of much current research in analytical chemistry. Of the measurement approaches under study, infrared spectroscopy offers the capability to implement a selective, nondestructive analysis of a variety of chemical samples. In an automated analysis, current infrared instrumentation produces volumes of data necessitating the development of data processing techniques to extract useful information. In this dissertation, automated analysis methods for qualitative and quantitative applications of infrared spectroscopy are explored. These methods are directed to environmental remote sensing and clinical applications and focus on both single-point measurements with Fourier transform infrared (FT-IR) instrumentation and imaging measurements performed with a multispectral line scanner. The developed methodology is directed to isolating the analyte signature from the data for use in a qualitative determination of analyte presence or in a quantitative measurement of analyte amount. A study was conducted to find optimal parameters for generating finite impulse response matrix (FIRM) digital filters for use in isolating analyte signals directly from FT-IR interferogram data. The filter design protocols established in this study are used to generate filters for quantitative and qualitative applications. Filters are designed to extract the glucose signal from a complex simulated biological matrix with severely overlapped spectra. Digital filters are also developed to isolate ammonia signals for use with pattern recognition techniques for the remote detection of ammonia in heated plumes from stack emissions. Analysis techniques are also developed for use with data from an infrared multispectral imaging system, designed to detect chemical plumes from stack emissions. These plumes are viewed from above by installing the imaging system on an aircraft platform. The alpha residual method is used to remove temperature effects from the images and thereby simplify the detection of chemical signatures. Pattern recognition methods are used to develop automated classifiers for detecting ethanol plumes from a controlled release experiment and methanol from an industrial facility.
Bibliographical Information:


School:Ohio University

School Location:USA - Ohio

Source Type:Master's Thesis

Keywords:chemistry and biochemistry


Date of Publication:01/01/2002

© 2009 All Rights Reserved.