| Summary: | This work presents investigations into the use of the near-infrared (NIR) signals to interrogate, detect and image specific chemical compounds of interest in a security screening application, including when such compounds are hidden behind single layers of clothing fabric. In an initial set of experiments, the mechanisms governing the interaction of NIR signals with clothing fabrics and similar materials has been studied, in order to account for the influence of fabric layers when detecting hidden chemicals. Throughout the rest of the work, NIR spectroscopy has been used as a means to perform qualitative and quantitative analysis, in order to detect the presence of chemicals, and quantify the concentration in aqueous solution of liquids. It has been shown that, while the compounds can be identified on the basis of the characteristic features that appear in the relevant NIR spectra, the origin and nature of these spectra necessitate that such identification be performed with a chemometricsbased approach. Accordingly, multivariate calibration models based on neural networks and partial least squares regression (PLSR) have been developed to perform the requisite analyses. Results of calibration and testing with a range of data are reported. In order to facilitate operation in practical security screening, the development and testing of a software-based lock-in amplifier is reported, as a mean to enhance the signal-to-noise ratio (SNR) of the spectral data. It is shown that the amplifier can process up to 40 wavelength channels in parallel, to extract the spectral data buried in noise in each channel. Hence, with the SNR of the input signal set as low as -60 dB (by introducing software-generated additive white noise in the spectra), adequate noise suppression has been obtained, allowing the resulting spectral data to be used for requisite chemical detection. Finally, an integrated spectroscopic imaging application is developed to perform twodimensional cross-sectional scans of chemical samples, carry out lock-in amplification of the recorded intensity spectra, and plot the results of neural network-based chemical detection in the form of intensity images colour-coded to depict the presence of the pertinent chemicals at the scanned coordinates.
|
|---|
