Search

Reflectance (emittance) spectroscopy, especially at infrared wavelengths, continues to grow in utility as an analytical technique for contact, standoff, and remote sensing. The reflectance spectra of solids, however, are complex, depending on many parameters, even for the same material. Granule or powder particle size, crystal morphology, layer thickness, and substrate material all affect the spectral distribution of reflected light. However, such phenomena can all be modeled if the optical constants n(ν) and k(ν) are available. If the quantitative absorption coefficient K(ν) is known, the k value can be obtained via the relation k(ν) = 2.303K(ν)/4πν. The absorption coefficient can in turn be derived from a simple KBrpellet infrared absorption measurement, provided the pellet mass ratio is prepared quantitatively. The method requires the pellet’s mass and diameter, along with the analyte mass fraction and density. In this paper we demonstrate the requisite experimental details in preparing the pellets, as well as methods to reduce light scattering in order to obtain more quantitative values. Theoretical methods to derive the related optical constants will also be detailed, in particular the assumptions used to obtain the scalar refractive index n. Ideally, this value is known or measured separately, but in some cases we have found that it can be approximated (first approximation) for most organic chemicals by n~1.5 at the shortest wavelength. The results are presented for a couple of species.

Trace particle quantities of explosives left behind by those handling explosives materials present an opportunity to identify both the handlers, secondary handlers and the objects they have contacted. Understanding the nature of these particles is critical for tailoring optical detection strategies as well as non-optical contact harvesting methods. We are working towards developing a model to understand and quantify the nature of particles transferred from the hands to different substrate surfaces. In this preliminary paper we report on a newly developed finger test-bed to produce a robotically controlled series of fingerprints, with an artificial finger designed to mimic the physicochemical properties of the human finger. In an initial set of experiments, we examine the effect of a range of applied forces, the effect of a range of initial particle sizes, and the serial print number on the deposited mass and deposited particle sizes, for a surrogate explosive loaded as particles on gloved fingers which are subsequently pressed against a set of clean glass slides.

The results from infrared backscatter imaging spectroscopy on a mobile platform for stand-off detection of trace amounts of explosive materials on relevant substrates are presented. This technique utilizes an array of tunable infrared quantum cascade lasers to illuminate targets. The spectral range of the QCL system spans from 6 - 11 μm, which enables excitation of a wide variety of absorption bands present in analytes of interest. Targets are prepared by sieving particles through a 20 μm mesh onto substrates to simulate relevant qualities (particle size, fill factor, and mass loading) expected of real world targets. The backscatter signal from targets is collected with an IR focal plane array. This information is stored in a hyperspectral image cube to allow for post processing in a detection algorithm. We demonstrate the selectivity and sensitivity of the discussed technique down to the nanogram level for RDX and PETN on glass. Spectra are generated by extracting the signal from small regions of interest to simulate targets with miniscule coverage areas. Preliminary comparison of backscatter data with simulated data from a model that incorporates particle size, mass loading, and substrate response show good agreement. Confusant agents, such as sand, are introduced to the targets loaded with analyte to illustrate the selectivity of this technique. The results of these studies are presented, along with future improvements to the technique.

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.