Abstract
Spectroscopic analysis in the Terahertz frequency range, providing characteristic "signatures" for explosive and non-explosive materials, is proposed as an efficient and powerful tool for explosive identification. It is demonstrated that spectral responses of materials can be used as fingerprints that distinguish cyclotrimethylenetrinitramine (RDX) from other materials even with simple detectors and a limited number of available frequencies. Detection is performed using a modified least squares approach and multilayer perceptrons that operate on smoothed reflectance spectra. The performance of the detectors is evaluated through application to spectra of RDX and several common materials. A Receiver Operating Characteristic curve analysis demonstrates that our detectors exhibit the desirable properties of high probability of detection and low probability of false alarm.
Original language | English (US) |
---|---|
Pages (from-to) | 1171-1181 |
Number of pages | 11 |
Journal | Journal of Infrared, Millimeter, and Terahertz Waves |
Volume | 31 |
Issue number | 10 |
DOIs | |
State | Published - Oct 2010 |
All Science Journal Classification (ASJC) codes
- Radiation
- Instrumentation
- Condensed Matter Physics
- Electrical and Electronic Engineering
Keywords
- Explosive detection
- Neural networks
- ROC curves
- THz spectroscopy