This PDF file contains the front matter associated with SPIE Proceedings Volume 12541, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

Most reported measurement efforts for visualizing gaseous exposure signatures aim to detect and analyze continuous releases of volatile chemicals. Recently, we became particularly interested in characterizing short-time explosive releases of chemical substances. To perform such experiments, we pursued the construction of a suitable device that generates appropriate short-time events in a reproducible manner. This device, which we refer to as an aerosol bomb, allows the controlled release of liquids from 10 to 200 mL within a timeframe of one to two seconds after being pressurized up to 80 bar. Furthermore, different spray profiles and, thus, different cloud shapes can be created using customized spray nozzles. These short-time chemical exposures, however, proved challenging to visualize by video recordings as dilution and volatilization led to the rapid disappearance of visible cloud shapes. Therefore, we utilized a dual setup of passive infrared (IR) Focal Plane Arrays to detect and identify these lower concentrations of chemicals. In preceding studies, we have already shown the application of an IR focal plane array detector for hyperspectral recording and analysis of measurement fields of various sizes with 128 x 128 pixels in a time grid of two seconds. After connecting two hyperspectral imaging measurement systems into a combined dual setup, we conducted a three-dimensional (3-D) characterization of short-time chemical exposures, whereby 3-D imaging is realized by intersecting beams of IR waves.

Rapid and selective detection of persistent, highly toxic liquids, such as low volatile chemical warfare agents (CWA) continues to be a desired capability, e.g. by first responders and military personnel. An already proven general technique for detection of bulk material is Raman spectroscopy. Utilizing UV excitation wavelengths offers advantages such as separation from fluorescence, solar blindness and a high Raman cross-section. In the UV region, however, there is usually a rapid decrease in penetration depth for most liquids as the wavelength becomes shorter, generally resulting in that a smaller volume is accessible for Raman scattered photons. While this effect is a drawback in terms of signal power at the detector it may also be beneficial as interfering light from the background surface can be strongly reduced or even absent. Herein, the use of Hadamard patterned (50% transmission) masks at the entrance plane of a spectrograph are investigated for the purpose of increasing the amount of Raman scattered light onto the detector compared to slit measurements. Decoded spectra from Hadamard measurements on scenes containing hazardous material, such as low volatile CWA and simulant chemicals, are compared with slit measurements.

Physical Sciences Inc. has developed a standoff deep ultraviolet (DUV) Raman sensor for the detection of explosive residues. The sensor is based on a solid-state DUV excitation source coupled with a Spatial Heterodyne Spectrometer receiver. The sensor measures Raman signals across a ~830–2680 cm^-1 spectral range from a 2.6 cm² interrogation area from a 1 m standoff in a single snapshot with a 17 cm^-1 spectral resolution. Acquired spectra are processed through an on-board deep learning spectral correlation algorithm that provides real-time target identification. Developmental testing of the sensor has been conducted in a laboratory environment against explosive simulants including potassium chlorate, ammonium nitrate, and urea in bulk form as well as residues deposited on various substrates including plastic, glass, and metals. These measurements have demonstrated the system’s ability to measure Raman spectra and identify targets in 1 to 120 seconds.

Raman spectra were measured using a novel experimental configuration. This configuration allows many of the difficulties associated with the collection of Raman spectra under proximal conditions to be mitigated. Large sample areas can be imaged into the detection system allowing low intensity (high power) excitation sources to be used while simultaneously avoiding sample degradation and multi-photon absorption effects. Such large detection areas allow high numbers of molecular scatters to be probed even with minimal penetration depth. Overlap of excitation and detection areas avoids alignment difficulties that plague conventional Raman while removing the focal plane making detection from near contact to proximal distances possible. Justification for the success of this optical configuration will be described along with supporting data.

Waveguide-enhanced Raman spectroscopy (WERS) using nanophotonic waveguides has been used to demonstrate the detection of vapor-phase chemicals and liquid-phase biomolecules in water. The technique benefits from the fabrication processes and tolerances of CMOS foundries, but successful foundry-based WERS photonic integrated circuits (PICs) have only been demonstrated using excitation wavelengths of 1064 nm and 785 nm. Foundry-based PICS are beginning to operate with low loss at visible wavelengths, and WERS is uniquely poised to take advantage of this capability. Raman scattering cross-sections scale as λ⁻⁴, so a visible WERS platform could enable increased sensitivity, decreased exposure times, and/or decreased laser powers. However, increased fluorescence, increased waveguide loss, and decreased feature sizes make WERS in the visible challenging. Here, we demonstrate WERS using 300-mm foundry-based fabrication (AIM Photonics) with 633 nm and 785 nm laser excitation. We also show the successful operation and integration of other required components for a compact WERS system operating in the visible, including edge-couplers and lattice filters.

We demonstrate a novel solid-state spectrometer employing a linear array of resonant cavity enhanced photodiodes (RCE-PDs) with a spatial chirp. By epitaxially grading the thicknesses of the distributed Bragg reflector mirrors, the chirp can cover a total bandwidth of ≥0.1 × λres where λres is the resonant wavelength. This new class of sensor is intended for analyzing IR absorption fingerprints and our group has already demonstrated conventional RCE-PDs between 2.2 – 7.8 μm. In theory the range between 1.55 and ~12 μm could be served using the same materials. This region covers important spectral fingerprints including chemical and pollutant gases, as well as threat agents including thiodiglycol and VX.

Alakai Defense Systems has recently developed what we believe is the first one-handed UV Raman sensor for standoff trace detection of chemicals, which we refer to as Argos. Argos is the higher performance version of lower cost SAFR sensor, since it has increased range and trace detection capability. Since it is lightweight, Argos can also be deployed on Unmanned Ground Vehicles (UGV’s) or Unmanned Aerial Vehicles (UAVs). Data is presented showing how Argos detection performance is essentially the same as Alakai’s proven trace detection sensor PRIED, however it is ~70% smaller and lighter.

We report results from a recent field experiment to test the validity of using physics-based synthetic infrared spectra to serve as endmembers in a spectral database targeted at chemical deposits. Specifically, the optical constants n and k, (the real and imaginary part of the refractive index) were used to first model infrared reflectance spectra for different thicknesses of chemical layers (e.g. acetaminophen, methylphosphonic acid – MPA, etc.) on various conducting and insulating substrates such as aluminum, wood, and glass. In the experimental portion of the research, thin films of the solid and liquid analytes were deposited onto such substrates to form micron-thick layers of the analytes at different thicknesses: Standoff data from an imaging instrument were then recorded and analyzed to not only identify the different analytes, but also quantify the layer/deposit thickness. To gauge success, the detection results using the synthetic data were compared to the results from laboratory hemispherical reflectance (HRF) spectra that were collected for the same sample planchets measured in the field via standoff methods. Preliminary results indicate good agreement between the synthetic reference data as compared to the lab-measured HRF data in terms of their ability to quantitatively reduce longwave infrared data. Specifically, modeled IR spectra for acetaminophen on an aluminum planchet at various thicknesses (1, 2, 5, 10, 15, and 20 μm) were synthesized and compared with standoff field reflectance data as well as HRF laboratory reflectance spectra for two samples: a 5.2 μm- and 12.8 μm-thick layer of acetaminophen on aluminum. Using a first-order approximation, analysis of the field data estimates the thicknesses of the samples to be 2 and 10 μm for the two samples, respectively, while the HRF laboratory data yields thickness estimates of between 5-10 μm and 10 μm, respectively. Both yield reasonable estimates, with the uncertainty most likely due to factors yet to be accounted for in the synthetic spectra such as light scattering.

Raman sensing and mapping techniques traditionally use a tightly focused laser beam to incite and collect Raman scattered photons. A large amount of energy is typically focused in a very small (micron-sized) area potentially resulting in photo-induced damage and can be not eye-safe. In addition, when using a focused-based laser system, scanning a large area is time consuming due to the small area of interrogation and must be done at a specific distance. Therefore, either prior knowledge of the sample location (in three dimensions) is necessary, or a smaller area must be scanned. In this work, we demonstrate a hand-held proximal Raman detection instrument that uses a non-focused laser beam to interrogate a larger area. This reduces the time it takes to map a surface and provides greater flexibility in targeting the area to interrogate. Herein, we show detection and mapping of explosives in two dimensions with this hand-held proximal Raman instrument as well distance dependence of this non-focused instrument with explosive materials.

Though thermoluminescent dosimeters (TLDs) are one of the most commonly used and well-known radiation detectors in the industry today, they do not typically provide real-time feedback. A radiation detector that could alert users in real-time and that is small enough to fit in a pocket would be very useful for those who work near sources of radiation or to counter radiation weapons. In this project, a compact, low-power, portable detector is being developed to provide real-time radiation detection and discrimination that allows the user to vary the threshold and sensitivity of detection based on the radiation intensity.

The concept of an autonomous rover system to perform maintenance, investigations, and data collection in remote or inaccessible locations has seen an increased demand recently. In this work, an autonomous rover is developed to detect radioactive contamination. The rover utilizes a gas tube radiation detector as an active sensing element and onboard modules to command and control the rover, such as a GNSS receiver, Autopilot controller, and a microcontroller as an onboard controller a communication module. The rover could be controlled by a human operator or autonomous control. In both cases, the operator would be far away from the scene. The rover has many potentially valuable applications, such as radiometric survey and mapping, locating survivors, or aiding in recovering victims after a CBRN disaster. This paper discusses the concept of operations and the design of the autonomous rover.

The progress in development of a dual-comb spectrometer for detection of traces of explosives at stand-off distances is reported. The spectral range of the spectrometer was extended to 1205-1305 cm-1, the stand-off distance was shortened to 0.5 m to access more potential use-cases, and the speckle contrast was decreased to 0.3%. Tests of the dual-comb spectrometer on RDX and PETN deposited on glass surfaces with a surface concentration of ~10 g/cm2 deposited using a sieving method will be presented and compared with the measurements carried out using a laboratory grade FTIR instrument.

A standoff trace chemical detection system (TCD) based on LWIR hyperspectral imaging has been developed to detect and identify a wide range of trace chemicals on a variety of shipping-related materials and surfaces. The system is able to perform the detection on a continuous stream of parcels as they move along a conveyor at speeds up to 400 ft/minute. Optical illumination is provided by miniature, widely tunable external-cavity quantum cascade lasers (EC-QCLs) and the reflected light is captured using a high-speed HgCdTe camera. The acquired images are combined into a hyperspectral image cube (i.e., hypercube) by taking into account the velocity of the parcels. The system was mounted on a gantry looking down onto the parcels. A motorized camera lens with fast focusing of <0.1 s enabled the system to focus on the top surface of each parcel. The resulting hypercubes were then analyzed using efficient image segmentation and detection algorithms to identify and map the trace surface chemicals. Applications include the detection of trace amounts of explosives and opioids on parcels in shipping, sorting, and customs screening scenarios.

Wires do not have linear structure to aid their detection when they are not lying straight. This paper proposes some features for curved wire detection from the images constructed by the data of a ground penetrating radar. We propose the application of a parabola to model a curved wire, where the detection confidence corresponds to how well the curved pattern in the image fits to a parabola. The processing involves projecting the 3-D GPR beamformed image onto the ground plane, applying the Canny edge detector to extract the edge points, and fitting the edge points to a parabola through a voting scheme as in the generalized Hough Transform. The features consist of the orientation angle, the fitted parabolic parameters and the fitting confidence. Some examples for the detection performance are illustrated.

Real-time analysis of data provides input for decision makers. However, in the battlefield, that could be the difference between life and death. Therefore, techniques must be developed that provide data in a way that can be reduced to real-time information. Hyperspectral data is often sought after as it provides spatial-spectral information but comes with a large computation cost. Real-time analysis of hyperspectral data is often difficult after an appreciable amount of time due to the volume of data that must be analyzed. However, commercial off-the-shelf instrumentation that normally outputs large hypercubes of information can be computationally managed in a way such that real-time processing is achievable at low levels of analyte. In this work, we show near-trace level anomaly detection of explosive precursors, explosives, and pharmaceutical surrogates on real-world surfaces using a commercial off-the-shelf instrument. The threat anomaly detection (ThreAD) algorithm that is employed uses a semi-supervised machine learning method to determine where the anomalous data (i.e. analyte) is present. This work will provide approximate limits of anomaly detection (LOADs) for some analyte/surface combinations in laboratory conditions.

A standoff trace chemical detection system to detect vehicle-borne threats was developed using a long-wave infrared (LWIR) microbolometer (MB) camera in combination with widely tunable external-cavity quantum cascade lasers. The system acquires hyperspectral images of the target surface’s reflectance in the LWIR portion of the “chemical fingerprint” band to allow for high-sensitivity detection and high-specificity identification of a wide range of surface chemicals. By using a MB camera, as opposed to more expensive alternatives, the system is targeted for applications that require small size and low cost. This talk describes the design and performance of the prototype.

The most accurate insight to how aerosolized material responds to UV radiation is obtained by performing experiments on freely suspended particles, absent from the shadowing that deposition on a surface may impose. For this purpose we have developed a linear electrodynamic particle trap to confine suspended particles using a contact-free technique. The trap allows us to challenge and study aerosols under controlled environmental parameters such as temperature, humidity and radiation exposure. We present the results of a quantitative study on the changes in viability of Bacillus anthracis Sterne strain spores confined within this trap and illuminated by either simulated sunlight or a UV light source at 253.7 nm. Up to 500 same-size particles, (that is, consisting of approximately the same number of spores), were created from a droplet-on-demand injector, trapped and irradiated with varying time scales. Illumination times ranged from 5 to 300 seconds with a maximum fluence of 500 J/m2 using the UV source, and particle clusters containing as little as 1 up to as many as 55 spores were used. As will be discussed, the viability of spores decreased as total fluence increased as expected, and for the same fluence, viability improved as the number of spores in each particle increased.

Alakai Defense Systems has developed several standoff ultra-violet (UV) Raman systems over the years to enable detection of hazardous chemicals from a safe distance. These systems have traditionally used classical non-machine-learning-based algorithms, but Alakai together with its partner Systems & Technology Research (STR) are currently developing the Agnostic Machine learning Platform for Spectroscopy (AMPS). AMPS, implemented using PyTorch, automatically creates and optimizes tailored one-dimensional (1D) convolutional neural networks (CNN) when trained on simulated or measured data. Several emerging and novel techniques, including advanced domain adaptation approaches, have been implemented to increase model robustness and minimize training data requirements. While the created models are optimized for a specific modality, AMPS itself is agnostic—it can be used for any spectroscopic modality that produces 1D spectra. AMPS has shown promising results for long-wave infrared (LWIR) reflectance spectroscopy as well as UV and near-infrared (NIR) Raman. This talk will focus on AMPS models created using both simulated UV Raman data as well as measured UV Raman data taken with Alakai’s Portable Raman Improvised Explosives Detection (PRIED) system. Performance between AMPS and Alakai’s legacy algorithms will be compared.

A machine learning based approach has been developed to classify Raman spectroscopic data. The algorithm is based on a one dimensional neural network (1D-CNN) architecture which is trained with synthetic data that can incorporate sensor specific characteristics such as spectral range, spectral resolution and noise. The synthetic spectra are based on high SNR measurements which are then augmented by mixing target and background signatures. The CNN is trained to consider target representations in the presence of certain background materials including glass and HDPE. These additional target representations allow the CNN to make detections for materials taken through a container. Within this paper the performance of CNNs trained for Raman sensor systems has been evaluated using real data collected using the ThermoFisher FirstDefender. The evaluation data consists of various target chemicals (including explosives) and interferents (including household materials) collected through glass and plastic vials. The data was acquired with a controlled range of collection settings, including integration time and laser power, available on the unit. The performance of the 1D-CNN approach has demonstrated high classification accuracies, high probability of detection and low false alarm rates. Specifically, these metrics have been calculated as a function of signal to noise ratio. Additionally, a sensitivity analysis was conducted using an acetonitrile standard diluted in water which demonstrates the CNN’s capability of detecting all dilutions of acetonitrile down to weight concentrations of <1%. This sensitivity analysis was mirrored using a mixture of potassium chlorate and Vaseline. The CNN demonstrated detections down to 10% by weight of potassium chlorate.

Traditional vapor sensing based on vibrational spectroscopy methods employs Fourier-transform infrared (FTIR) spectroscopy, which can produce reliable identification of ppm-level chemical vapors; however, chemical warfare agent poses a threat down to ppb and lower levels. Novel cavity ring-down spectroscopy-based sensors offer a possible path toward combining the high amount of spectral fingerprint data available from traditional IR methods, and the sensitivity of higher-sensitivity technologies, such as nanomaterial arrays and ion-mobility spectroscopy (IMS) which lack high amounts of data which can be used to uniquely identify molecules and are still plagued by high false-alarm rates and low selectivity. Cavity ring-down expands the effective path length of a gas analyzer by orders of magnitude by reflecting the IR light back and forth across an analyzer cavity. By characterizing the loss of light with each bounce by understanding the reflectivity of the mirrors employed at each end of the cavity, a characteristic "ring-down" time of the reflected light through a gas medium can be measured. Then, as analyte is introduced into the cavity, the ring-down times at each wavelength are shifted as a function of the absorbance of the analyte. Comparison of the measured vs. expected ring-down times can be interpreted to produce an IR absorption spectrum of the analyte. The effective pathlength of on order of kilometers allows for extremely high sensitivity beyond the capability of modern FTIR-based analyzers; however, this sensitivity comes at the expense of easier-to-saturate detectors as well. Additionally, cavity ring-down systems have already demonstrated the ability to measure absorption of aerosols both solid and liquid, filling a major gap left by FTIR vapor analyzers and opening the possibility of bioaerosol detection. We present a comparison of the two technologies and where they complement as well as fill in gaps in detection capabilities and offer a path forward for future generations of CB detection.

Colorimetric chemical sensors are some of the simplest and low-cost sensors available today. Advancements in small, chip-based optical readout packages, coupled with low-power electronics and wireless communications modules, have allowed for the development of a hands-free colorimetry-based chemical sensor that effectively acts as self-reading M8 paper that responds to vapor-phase chemical threats. We have designed a prototype that can be worn by the user and operate autonomously to provide individual-level chemical threat early notice. By integrating these sensors into a communications network, opportunities exist to create team-scale mesh chemical sensor networks. These prototypes have been tested against toxic industrial chemical (TIC) vapors in the lab and have undergone surety testing against select chemical agent threats. In an effort to extend the utility of these chemical sensors, we have recently started a new effort to integrate the sensor package into new form factors for new tasks. Examples of these include a throwable ball sensor for proximal remote point sensing (in the field or probing an enclosed space) and an unobtrusive sensor placed around a fixed site to monitor for upwind chemical releases.

An ongoing weapons of mass destruction (WMD) threat is the aerosolization of low-volatility chemical weapons agents (CWAs) such as the V-series nerve agents, Novichoks [1], and pharmaceutical-based agents (PBAs) e.g. fentanyl and other high-potency opioids [2]. These materials are liquids or solids at room temperature and generally have extremely low vapor pressures, rendering them difficult, if not impossible, to detect with conventional vapor-only detection systems. If these materials are aerosolized, they can disperse over broad geographical areas posing an immediate risk to individuals encountering them while airborne. Upon settling, a wide swath of persistent agent contamination may be left behind, posing an on-going surface contact threat or re-aerosolization risk. Field detection of toxic aerosols remains a significant challenge for chemical detectors. Most of the existing hand-held chemical detection technologies are not equipped to screen ambient air samples for the presence of aerosolized threats, leaving a capability gap among the most widely deployed point sensing technologies. Recently, 908 Devices released an aerosol module (the ‘Aero’) that is compatible with the MX908 handheld mass spectrometer. The combination of the aerosol module with portable mass spectrometry enables the detection of aerosolized threats within seconds at concentration levels that enable warfighters and first responders to take protective action quickly and minimize the impact from this alternative threat class.

Transitioning a technical method from the laboratory bench to the field is a challenge. Initially, the method needs to fill a technical gap to a degree that a warfighter or first responder would find additional hardware and training worth the logistical burden. Second, the method should be robust to minor deviations and interferents. Finally, the resultant end point must be easily read, understood, and provide actionable information to the user. Accomplishing all these steps is key to demonstrating the value of scientific research to the warfighter and delivering a valuable tool. Recent efforts have been focused on developing methods for easy and robust trace analyte collection and portable sample identification. The analytes of interest include explosives, pharmaceutical based agents, and drugs of abuse. The collection method involves paper modified with pressure-sensitive adhesives, i.e. yellow sticky notes, to sample various types of solid, porous, and environmental surfaces. Threat identification is performed directly from the collection substrates by mass spectrometric instrumentation with tandem capabilities to identify TNT, RDX, and HMX. The surface limits of detection (LODs) of the method ranged from sub to low microgram range. An analysis mode was created that would display a green light/red light if a sample was negative/positive, respectively, for a threat. This provides an easy-to-read, actionable result while saving the analytical spectra for future review. Finally, this methodology was combined with portable Raman analysis to provide both primary and confirmatory identification of fentanyl in simulated samples and TNT in samples both collected and analyzed in an austere location.

Detection of analytes deposited on surfaces is crucial for many applications: Development of methods to prepare thin layers (e.g. ~5 to 100 μm) is important for both system design and field studies. In this work, solid and liquid analytes were deposited on painted and bare substrates including aluminum, glass, plastic, and concrete using an ExactaCoat ultrasonic spray coater. Laboratory hemispherical reflectance (HRF) spectra were collected for samples with different layer thicknesses so as to characterize both the composition and layer thickness. Preliminary results demonstrate that to prepare homogenous layers on surfaces, parameters such as substrate type, analyte solubility, vapor pressure, paint color, surface porosity, and surface roughness are all important. Liquid chemicals posed several issues during deposition: Diisopropyl methyl phosphonate evaporated from surfaces more quickly than the other chemicals and was thus not detected in the HRF experiments. Less volatile liquids, such as tributylphosphate, remained on the surface for the duration of the test, but a uniform layer thickness could not be obtained as the liquid pooled to one side when mounted at an angle. The deposition of solids (e.g., acetaminophen, caffeine and methylphosphonic acid) from volatile solvents such as chloroform also proved problematic due to streaking caused by rapid solvent evaporation. Solids deposited from ethanol, however, worked well on bare substrates. For most samples plotting the integrated infrared band strength vs. surface thicknesses showed a linear relationship, confirming that the surface loading can be controlled by programming the concentration and the number of passes on the ultrasonic sprayer.

Ellipse feature appears in the cross-section of a wire, and it can complement the line feature to improve wire detection. The previously proposed ellipse feature extraction method requires a wire to be placed parallel or perpendicular to the cross-track. It is, however, difficult to guarantee in practice. This work advances ellipse feature extraction so that we will be able to obtain the feature regardless of the orientation of a wire. The method first applies the Hough Transform to the surface projection of an object image from the ground penetrating radar, and then rotates the 3-D data image according to the orientation angle from the Hough Transform to align with the cross-track, before the extraction of ellipse feature. The proposed method is quite effective and provides high quality ellipse feature to aid the detection of wire in any orientation.

By using IMU, it is aimed to correct the distortions in the vehicle-mounted stepped frequency continuous wave (SFCW) ground penetrating radar (GPR) systems, which occur as a result of the antennas not maintaining their position relative to the surface due to the roughness and the random motion of the antennas within a certain stroke with respect to time. In this context, detrending and numerical integration algorithms were used for motion compensation with real GPR data, which was taken on the vehicle. These algorithms used to obtain instantaneous displacement of the random motion of the antenna relative to the ground. Velocity component in the z direction was filtered with the high-pass filter and displacement was calculated with the numerical integration. Then, motion compensation was made by adding the obtained displacement amount to the real GPR data as a phase. Comparing the motion-compensated and non-motion-compensated conditions, it was seen that the real GPR data was improved dramatically as a result of the operations. Furthermore, it was determined that the real GPR data with the algorithm applied, gives better results in terms of target detection.