explosives-terahertz-imagin-quantification

Loading ...
terahertz
Science
Abstract—Chemometric analysis was applied on terahertz 
absorbance 3D images, in transmission. The goal is to 
automatically discriminate some explosives on images and 
quantify mixtures of RDX/PETN in the frequency range of 0.2 – 
3 THz. Partial Least Square (PLS) was applied on THz 
absorbance multispectral images to quantify individual product 
inside pure samples and mixtures at each pixel on the image. 
Then the best score obtained is used to display the samples’ 
images and provide the optimal frequencies combination for 
recognition purpose. 
I. INTRODUCTION 
erahertz spectral-imaging is an imaging technique that 
utilizes the wave in the terahertz region of the 
electromagnetic spectrum (3-0.03 mm; 0.1-10 THz) [1]. It 
was investigated in different fields such as biology, 
pharmaceutical, security fields, etc., and allows getting 
physical and chemical information about the inspected 
samples by getting the dielectric response over several 
decades of frequencies. 
On the other hand, chemometrics technics showed powerful 
capability in data analysis for discrimination, identification or 
quantitation. As well as in terahertz science, chemometrics 
was successfully applied on spectroscopy [2] and on imaging 
[3] for various purposes. 
Partial least square (PLS) is one of the chemometrics 
multivariate analyses used to get chemical information on 
multispectral data and to predict the amount of products in 
samples. Different studies demonstrated the good ability of 
PLS applied on THz spectra for quantification of different 
chemical substances [4]. 
In the context of transportation security and in the aim of 
not only identify some explosives but also to predict their 
exact amount in the terahertz domain, we applied PLS 
algorithm on multispectral terahertz images. This study was 
made 
in 
order 
to 
quantify 
two 
explosives 
cyclotrimethylenetrinitramine 
(RDX) 
and 
pentaerythritoltetranitrate (PETN) in pure state and in mixtures 
on each pixel of THz multispectral images. 
II. RESULTS 
Three different pellets of 400 mg total weight and 13 mm 
diameter were imaged for this study. The first and second 
pellets contain 80 mg RDX and PETN respectively. The third 
pellet contains 40 mg RDX and 40 mg PETN. Also, 36 pellets 
of pure RDX, pure PETN or mixtures, with different amounts 
ranging from 2 mg to 100 mg. For all the pellets, polyethylene 
(PE) was used as a binder. Terahertz measurements were 
obtained from time domain spectral imager, TPS spectra 3000 
from Teraview, working in transmission mode. The selected 
frequency band is 0.2-3 THz. Then, a Fast Fourier Transform 
leads to extract the absorbance spectra using PE measurements 
as reference. The resulted image has a size of 39*13 mm2 
which correspond to a 66*22 pixels. The 3D matrix of data 
was transformed in a 2D matrix PxF, where P corresponds to 
1452 pixels in the image and F corresponds to 486 
frequencies. It should be mentioned that each F vector 
corresponds to an entire image. Then, a PLS model was build 
where 622 absorbance spectra of the 36 pellets was used to 
calibrate the model, and the 2D matrix, unknown for the 
model, was used in order to predict the amount of RDX and 
PETN on each pixel of the 1452 on the entire image. The 
results show good ability of the model in prediction, with a 
RMSE lower than 4 mg for the entire 400 mg pellet. We will 
also present our model, the results in term of scores, loadings, 
root mean square error (RMSE) and the spatial repartition of 
predicted values on the image. 
 
Fig 1: (a) Photo of three different pellets of 400 mg total weight constituted 
respectively of 80 mg RDX, 40 mg RDX/40 mg PETN and 80 mg PETN, (b) 
image of the predicted pixels: blue, orange and red correspond to 80 mg RDX, 
40 mg RDX/40 mg PETN and 80 mg PETN respectively with an RMSE less 
then 4mg. 
III. SUMMARY 
In this study we showed the ability of PLS analysis applied 
on THz multispectral absorbance images, in quantification of 
RDX and PETN explosives inside mixtures, not only from one 
THz spectrum measurement, but on an entire multispectral 
image. So with this procedure, we can detect, quantify and 
display explosives hidden inside pellets. 
REFERENCES 
[1] Y. Watanabe, K. Kawase, T. Ikari, H. Ito, Y. Ishikawa, and H. Minamide, 
“Component spatial pattern analysis of chemicals using terahertz 
spectroscopic imaging,” Appl. Phys. Lett., vol. 83, no. 2003, pp. 800–802, 
2003. 
[2] Y. Chen, Y. Ma, Z. Lu, B. Peng, and Q. Chen, “Quantitative analysis of 
terahertz spectra for illicit drugs using adaptive-range micro-genetic 
algorithm,” J. Appl. Phys., vol. 110, no. 2011, 2011. 
[3] Y. Shen, P. Taday, D. Newnham, and M. Pepper, “Chemical mapping 
using reflection terahertz pulsed imaging,” vol. 5727, pp. 24–31, 2005. 
[4] J. El Haddad, F. De Miollis, J. Bou Sleiman, L. Canioni, P. Mounaix, and 
B. Bousquet, “Chemometrics applied to quantitative analysis of ternary 
mixtures by terahertz spectroscopy,” Anal. Chem., vol. 86, no. 10, pp. 
4927–4933, 2014.  
 
J. Bou Sleiman1, J.B. Perraud1, B. Bousquet2, N. Palka3, J.P. Guillet1, and P. Mounaix1 
1Bordeaux University, IMS, CNRS UMR 5218, 33405 Talence, France 
2Bordeaux University, CELIA, CNRS UMR 5107, 33405 Talence, France 
3Military University of Technology, Institute of Optoelectronics, Kaliskiego 2, Warsaw 00-908, Poland 
Chemical imaging and quantification of RDX/PETN mixtures by PLS 
applied on terahertz time-domain spectroscopy 
T