Search

We are developing algorithms to identify chemicals of interest by their diffuse infrared (IR) reflectance signatures when they are deposited as particles on surfaces. For capturing the signatures themselves, we are developing a cart-based mobile system for the detection of trace explosives on surfaces by active infrared (IR) backscatter hyperspectral imaging (HSI). We refer to this technology as Infrared Backscatter Imaging Spectroscopy (IBIS). A wavelength tunable multi-chip infrared quantum cascade laser (QCL) is used to interrogate a surface while an MCT focal plane array (FPA) collects backscattered images to comprise a hyperspectral image (HSI) cube. The HSI cube is processed and the extracted spectral information is fed into an algorithm to detect and identify chemical traces. The algorithm utilizes a convolutional neural network (CNN) that has been pre-trained on synthetic diffuse reflectance spectra. In this manuscript, we present an approach to generate large libraries of synthetic infrared reflectance spectra for use in training and testing the CNN. We demonstrate advancements in the number of analytes, a method to generate synthetic substrate spectra, and the benefits of subtracting the substrate “background” to train and test the CNN on the resulting differential spectra.

This study describes inverse spectral analysis of diffuse reflectance for surface-distributed material particles on substrates. In particular, an algorithm for extraction of target spectral features for surface-distributed materials of specified dielectric response. This algorithm is based on diffuse-reflectance theory and linear combinations of basis functions representing response characteristics of different types of scattering processes. The basis functions are constructed using absorbance functions and analytical models of Mie-type scattering. Prototype inverse spectral analysis of diffuse reflectance for surface-distributed explosive particles on substrates are described, which demonstrate characteristics of the algorithm.

This study describes a methodogy for spectrum-feature extraction from diffuse reflectance for distributions of materials on substrates, which is based on diffuse-reflectance theory and phenomenological multiplicative-factor decomposition of reflectance functions. Specifically, this methodology entails feature-extraction using reflectance-spectrum normalization with respect to phenomenological backgrounds. A mathematical analysis of the feature-extraction methodology with respect to its formulation is presented. In addition, results of inverse analyses demonstrating application of the methodology are described.

We present a cart-based system based on infrared backscatter imaging spectroscopy (IBIS) for detecting and analyzing trace amounts of hazardous materials as particles on solid substrates. A system comprising four quantum cascade lasers rapidly scans through the mid-LWIR (6 μm – 11 μm) wavelength range to illuminate samples containing target analytes. The infrared backscatter signal is collected as a series of images to form a hyperspectral image cube. Each image is collected at a specified excitation wavelength using a liquid nitrogen cooled MCT focal plane array. The experimental results of this cart-based infrared illumination and backscatter detection are presented. Results compare imaged spectra over a range of different wavelength tuning speeds and different combinations of substrates and analytes. Camera frames are collected while the laser is sweeping through its wavelength range. A single complete analysis can be completed in less than 1 second. In every camera frame, each pixel of the 128x128 pixel camera array produces an individual intensity. These frames are then binned and assigned a discrete wavelength in steps, typically 0.01 μm, to produce a spectrum over 6 – 11 μm for each camera pixel. Target samples are prepared by sieving particles or by a dry transfer technique, to mimic particle size distributions associated with real world threats at trace levels, for explosives and illicit drugs on relevant substrates.

Far-field infrared (IR) spectroscopy techniques, such as ellipsometry and FTIR, can yield extremely accurate measurements of the optical constants (n,k) which characterize the electronic and lattice-structural degrees of freedom in novel materials and devices. However, large systematic uncertainty has plagued extensions of these techniques to small length scales. A low uncertainty embodiment would enable high quality studies of, for example, the newest correlated condensed matter systems - where typically only a small sample is available early on. Here we present mature far-field IR microscope. Most notably, an asymmetric interferometer is used to directly measure the infrared reflectance amplitude and absolute phase shift. The complex optical constants can then be extracted without the large uncertainty that arises with an amplitude measurement alone.

We are developing a cart-based mobile system for the detection of trace explosives on surfaces by active infrared (IR) backscatter hyperspectral imaging (HSI). We refer to this technology as Infrared Backscatter Imaging Spectroscopy (IBIS). A wavelength tunable multi-chip infrared quantum cascade laser (QCL) is used to interrogate a surface while an MCT focal plane array (FPA) collects backscattered images. The QCL tunes across the full wavelength range from 6 – 11 μm. Full 128 X 128 pixel frames from the FPA are collected at up to 1610 frames per second and comprise a hyperspectral image (HSI) cube. The HSI cube is processed and the extracted spectral information is fed into an algorithm to detect and identify traces of explosives. The algorithm utilizes a convolutional neural network (CNN) and has been pre-trained on synthetic diffuse reflectance spectra. In this manuscript, we present backscatter data and hyperspectral image mapping from a car panel substrate deposited with traces of the explosive RDX. We have used a mask to restrict the RDX analyte deposition to small 4 mm diameter areas. The results presented here were measured at 1 meter standoff.

The use of rapid scanning quantum cascade lasers in the detection of trace amounts of explosive materials is presented. This technique, infrared backscatter imaging spectroscopy (IBIS), utilizes an array of quick tuning infrared quantum cascade lasers (QCLs) to illuminate targets with mid-IR light, 6 – 11 μm in wavelength, to perform measurements in less than one second. The backscattered signal from targets is collected with a liquid nitrogen cooled MCT focal plane array. This information is stored in a hyperspectral image cube which is then run through a detection algorithm which has been trained on synthetic reflectance spectra of analytes of interest. We discuss the experimental parameters used with the QCLs and the focal plane array to generate and collect the infrared backscatter signal. The performance of the fast scanning QCL is presented in detail along with the experimental protocol used to collect high quality data from targets at proximal standoff distance. Camera frames are collected as the laser wavelength is swept and then are binned and assigned discrete wavelength steps. Spectra are extracted from the binned frames on a pixel by pixel basis. When run at full frame imaging, this results in over 16,000 individual spectra.

The ability to rapidly detect hazardous airborne chemicals with high fidelity in a single point-detection system remains a significant challenge in a complex chemical background. Traditional Gas chromatography (GC) can significantly augment most detection technologies by separating complex mixtures for high fidelity detection, but with the disadvantage of requiring detection at the end of the GC column which adds a time disadvantage for any decision making process. Microfabrication of GC columns has reduced device footprint and power consumption, but the end-of-column detection paradigm remains. We present a rapid detection concept of in-column detection by probing the GC stationary phase which is coated on an IR transparent column substrate. The optical evanescent field interactions in the mid-infrared spectral region (US. Patent# 9,599,567) allows analyte detection along the column without having to wait for complete elution. These spectral signatures, collected at different points along the column, are analyzed by an algorithm to quickly identify components in a complex mixture. We present results with an ATR-based system that uses a focused tunable quantum cascade laser beam directed by galvo mirrors at points along a molded micro-GC column whose base comprises an optically transparent material coated with the stationary phase.

Optically Super-resolved InfraRed Imaging micro-Spectroscopy (OSIRIS) permits the collection of fully broadband hyperspectral images at the maximum optical spatial resolution. Modulated long wavelength light is directed onto the sample, while a short wavelength probe beam senses the resultant modulation in local temperature. Thus, OSIRIS constitutes the full generalization of color vision to full spectral bandwidth with spatial resolutions up to the maximum achievable with far-field optics (λprobe/4NA). The non-contact nature of OSIRIS permits deep tomography and real-time imaging, while preserving the versatility and ease of use of optical microscopy.

This study describes parametric modeling of diffuse IR reflectance for sparsely surface-distributed particles of the explosive RDX. Diffuse IR spectra are modeled using a formulation that considers spectral features due to target-material reflectance, i.e., RDX, substrate reflectance and resonance scattering resulting from finite sizes of surface-distributed particles. The results of this study demonstrate an approach for parametric modeling of diffuse IR reflectance for sparsely surface-distributed particles. The mathematical formulation of this approach is that of a phenomenological scattering-matrix representation.

Threat chemicals such as explosives may persist on surfaces, enabling them to be detected by non-contact or standoff optical methods such as diffuse IR reflectance. However, due to particle size effects and optical coupling to the substrate, their IR spectral signatures will differ from laboratory reference measurements of bulk materials. This study presents an inverse analysis of diffuse IR reflectance from sparsely surface-distributed particles of the explosive PETN. A methodology using spectrum templets is applied for inverse analysis of measured spectra. The methodology is based on a generalization of extended multiplicative signal correction (EMSC). The results of this study demonstrate application of the inverse analysis methodology for extraction of spectral features for surface-distributed particles of specified dielectric response.

Machine learning based perception algorithms are increasingly being used for the development of autonomous navigation systems of self-driving vehicles. These vehicles are mainly designed to operate on structured roads or lanes and the ML algorithms are primarily used for functionalities such as object tracking, lane detection and semantic understanding. On the other hand, Autonomous/ Unmanned Ground Vehicles (UGV) being developed for military applications need to operate in unstructured, combat environment including diverse off-road terrain, inclement weather conditions, water hazards, GPS denied environment, smoke etc. Therefore, the perception algorithm requirements are different and have to be robust enough to account for several diverse terrain conditions and degradations in visual environment. In this paper, we present military-relevant requirements and challenges for scene perception that are not met by current state-of-the-art algorithms, and discuss potential strategies to address these capability gaps. We also present a survey of ML algorithms and datasets that could be employed to support maneuver of autonomous systems in complex terrains, focusing on techniques for (1) distributed scene perception using heterogeneous platforms, (2) computation in resource constrained environment (3) object detection in degraded visual imagery.

We are using active infrared (IR) spectroscopic imaging to detect trace explosives on surfaces at proximal distances up to a few meters. The technology comprises an IR quantum cascade laser (QCL) for illumination and an IR focal plane array (FPA) sensor to collect signal backscattered from surfaces of interest. By sweeping the wavelength of the QCL while collecting image frames with the FPA, we generate an active hyperspectral image (HSI) cube. The HSI cube contains both spatial and spectral information, where the spectrum of a pixel, or region of interest within the image, can be extracted and compared against a known threat library. These cubes are fed into a convolutional neural network (CNN) trained on purely synthetic data to identify chemicals in the field of view. The CNN identifies chemicals by their IR signature and identifies their location within the image.

Gas chromatography (GC) is a staple analytical technique used to separate chemical mixtures (analytes) prior to identification with a hyphenated technique, such as mass spectrometry or Fourier transform infrared (IR) spectroscopy. Traditionally, analytes elute through the GC separation column where they are detected when they exit. We have developed a technique to perform in situ IR spectroscopy during the process of separating the analytes along the GC column. This is achieved by spin coating the stationary phase onto a germanium prism and actively probing the stationary phase in an attenuated total reflectance configuration with a quantum cascade laser.. The GC column is formed by pressing a molded epoxy lid, with grooves that form the tubular column, onto the stationary phase coated prism.

Rapid scanning quantum cascade lasers are utilized in the detection of trace amounts of explosive materials. Infrared backscatter imaging spectroscopy employs a quick tuning infrared quantum cascade laser system to illuminate targets with mid-IR light, 6 – 11 μm in wavelength, and to perform spectroscopic measurements in less than one second. A narrow cone of the signal backscattered from targets at standoff distance is collected and imaged onto a liquid nitrogen cooled MCT focal plane array. This backscattered signal is processed into a hyperspectral image cube containing spectral and spatial information. The analysis of the experimental data measured with the system is discussed. This includes the processing of the raw camera frames (using signals from individual components of the system) into discrete wavelength bins, typically 0.01 μm in width. Spectra are generated by plotting the signal from regions of interest, typically clusters of adjacent pixels within the frames, as a function of the wavelength associated with the binned frames. These spectra are compared against the FTIR diffuse reflectance of the analytes on an equivalent substrate for identification. Methods to optimize signal to noise and produce identifications with high confidence are presented. For a single experiment, taking less than 1 second, with the camera running at full frame over 16,000 individual spectra are generated. Targets are prepared by sieving and also dry transfer to mimic real world threats, in trace amounts and on relevant substrates. Traces of explosives, as well as illicit drugs are investigated.

Optical (visible) microscopy has established itself as an invaluable tool in Materials Science, and perhaps the canonical technique in Biology. Unfortunately, an optical image provides extremely limited information regarding chemical composition. A great deal of effort has been spent to circumvent this limitation through the use chemical dyes - most notably fluorescent markers - to achieve contrast between different chemical species. Infrared spectroscopy is an ideal technique to chemically fingerprint materials. Specifically, infrared microspectroscopy holds great promise as a labelless technique to achieve chemically specific microscopy. Unfortunately, the long wavelengths of the infrared lead to low spatial resolution in infrared microscopy, on the order of several microns. Traditionally, this limitation has been circumvented via scanning probe techniques such as s-SNOM and AFM-IR. While the scanning probe techniques provide excellent resolution, their contact nature and low signal levels limit the speed at which images can be acquired. We have developed a new technique to collect infrared hyperspectral images below the IR diffraction limit. Optically Sensed photothermal InfraRed Imaging micro-Spectroscopy (OSIRIS) permits the construction of infrared images on a resolution limited by the wavelength of the probe beam. In this technique, an infrared laser is used to excite the sample, while a short wavelength probe beam senses the resultant change in temperature. With this technique, hyperspectral images can be acquired orders of magnitude faster than the scanning probe techniques. Furthermore, a confocal setup permits tomography, which is extremely limited in the scanning probe techniques due to their surface nature.