This PDF file contains the front matter associated with SPIE Proceedings Volume 11727, including the Title Page, Copyright information and Table of Contents

Renaissance at NGA: Presented at SPIE Defense + Commercial Sensing, 2021

This paper will focus on analyzing the Covariance Equalization (CE) change detection algorithm for hyperspectral data. We analyze its weaknesses and suggest a method to improve the algorithm using normalization.

In image processing, the Matched Filter algorithm uses the estimated covariance matrix to give each pixel a score based on the similarity between the pixel and the signature of the target. While using this target detection algorithm, false alarms are inevitable. In order to solve this problem, a method using an iterative process to produce a second covariance matrix which only uses the most likely false alarms was presented [6]. In this paper, we test this method, attempt to improve it, and expand on the cases in which it is the most effective. In all cases, the new method showed a decrease in false alarms, and in some cases a decrease of over 85%.

A uniformly most powerful algorithm is derived for the standoff spectral detection of absorptive gaseous plumes. No weak concentration approximation is invoked. Other methods then naturally follow for the detection and estimation of gas mixtures. The algorithms result from the modeling of spectral backgrounds with multivariate lognormal distributions.

Inspired by the comic faux-Latin aphorism veritas duo sigma" (truth at two sigmas), an approach to target detection is proposed based on a likelihood ratio test in which the unknown target strength is treated as known, with strength chosen to correspond to a minimal level of detectability. For Gaussian distributions, this strength typically corresponds to two or three sigmas. This detector is admissible, which means that there is no other detector that is uniformly superior to it. The simplicity of the veritas detector permits closed-form solutions to be derived for a variety of signal detection problems. In a series of numerical experiments, these simple detectors are compared to traditional detectors, such as the locally most powerful detector and the generalized likelihood ratio test detector.

In 2018 the IEEE P4001 working was formed from industry specialists to facilitate consistent use of terminology, characterisation methods and data structures. This talk is a progress report to inform the hyperspectral community of the status of the work to date, the interconnection with other standards and outline the roadmap towards completion of the first draft of the standard for voting in early 2022.

The dynamic range of a hyperspectral camera can have a strong impact on the quality and integrity of the collected spectra. For the image sensor chip alone, the dynamic range is the ratio between the saturation level and the noise floor, set by full well capacity and readout noise respectively. In a complete hyperspectral camera, the raw signal level varies with the wavelength-dependent detector quantum efficiency, optics transmission, and illumination, as well as with the optical bandwidth. In practice, different parts of the spectrum will tend to have the highest or lowest signal level, and will most easily reach the saturation level or noise floor, respectively. It is shown that to define a dynamic range for a hyperspectral camera in the reflective spectral domain, the wavelength dependence of the camera light collection efficiency must be taken into account. A more application-oriented dynamic range can be defined by additionally assuming a shape of the illumination spectrum representative of a particular application.

This paper describes the development of a specially designed multi-modal microscope system that provides hyperspectral imaging from 300 to 1,700nm with a spatial resolution as high as 1µm in each of its optical measurement modes, i.e. reflectance, transmission, luminescence, and polarisation. The microscope system is fully automated and can handle large area samples from forensic tape-lifts. Demonstration of performance using tape-lift evidence samples that were seeded and studied to mimic real-life crime evidence collections at scene will be shown

Many research libraries and museums hold unique or rare items on which historically significant text is no longer legible due to damage, deterioration, or erasure. Spectral imaging { the process of capturing images of objects in many colors or wavelengths of light, including parts of the electromagnetic spectrum that are not observable by humans but easily imaged by modern sensors (i.e., ultraviolet and infrared) { has become the go-to" solution for recovering obscured and illegible texts. Unfortunately, most of these imaging systems are very expensive and not intuitive to use. Additionally, software to process the captured images can be expensive, difficult to use, and require significant knowledge of image processing methods. To address the above issues, we have developed a low-cost multispectral imaging system with accompanying open-source software that librarians, curators, and scholars with limited budgets can use to recover obscured and illegible text in their collections. The developed system is easy to use, can be dismantled and transported with little effort, and produce good quality spectral images as well as accurate true color renderings for digitization if needed. The software was developed with simplicity and functionality in mind. Basic image processing to uncover lost text is easy to implement without special image processing knowledge. The system specifications, characterization and calibration procedures and results are discussed. Images captured from a 15th century palimpsested manuscript leaf are shown and results discussed.

Cryogenic cameras are an innovative alternative in the design of miniaturized infrared cameras using cryogenic detectors. In this presentation, we will apply this technology to design a snapshot multispectral camera for gas leak detection. A UAV compatible demonstrator in a commercial Detector Dewar Cooler Assembly (DDCA), called SIMAGAZ, has been made and tested in the TOTAL Anomaly Detection Initiatives (TADI) platform and the Esperce site of ONERA. The TADI infrastructure manages monitored gas leaks at flowrates from 0.1 g/s to 300 g/s and hosts remote sensors to test them in three scenarios: crisis-management, safety monitoring, and environmental monitoring. In Esperce site, first UAV flights with SIMAGAZ were performed. We demonstrate the ability to detect and quantify in real time the origin of methane gas leak, the flowrate and the volume of the plume with SIMAGAZ on ground or from a UAV. The core camera weights around 1kg, for around 1L footprint and a power consumption of 10W at the cooling steady state. Results from TADI and Esperce campaigns will be presented.

The Moderate-resolution Imaging Spectroradiometer (MODIS) instruments are on-board the Aqua and Terra spacecraft, launched in 2002 and 1999, respectively. Since beginning operation, they have continued to collect valuable remote sensing data of the Earth in 36 spectral bands ranging in wavelengths from 0.41 to 14.5 μm. The Level 1B (L1B) algorithm produces calibrated top of the atmosphere (TOA) radiances for each Earth-view (EV) pixel and calibrated TOA reflectances for the reflective solar bands using geo-located, uncalibrated instrument data and calibration look up tables (LUTs) produced regularly by the MODIS Characterization Support Team (MCST). The L1B algorithm also calculates an uncertainty value for each EV pixel. The calibrated radiance and reflectance products are used to generate higher-level science products. A separate L1B code version is maintained for both MODIS instruments so that sensor specific issues can be handled individually. The current L1B algorithm version produces the Collection 6.1 (C6.1) products and was released in 2017. An overview of the C6.1 algorithm is provided together with improvements made since its release. Also discussed briefly are the planned improvements in the Collection 7 L1B algorithm.

The MODIS imaging spectroradiometer instruments on-board NASA's Terra and Aqua satellites have 20 reflective solar bands (RSB) covering a wavelength range from 400 to 2200 nm. Radiance is calculated from processing raw signals with background, temperature, and electronic contamination corrected. Measured gain is calibrated with a fully Sunlit solar diffuser (SD) at a stable radiance level, considering a slowly changing SD reflectance degradation. These measurements provide time-dependent gain adjustment factors, and the calibration assumes a linear response for each band and detector. Hence, an analysis of the dependence on different radiance levels is warranted. The MODIS design has no mechanism for varying radiance levels, except for an attenuator screen. However, it has been in static configuration for Terra since mid-2003. An external source of radiance attenuation can be utilized during solar eclipse events, while maintaining high stability and accuracy of solar calibration standards. Due to its long mission lifetime, Terra has seen several Sun-Moon near-conjunction events when it coincides with the orbit path where the SD is directly illuminated. As of August 2020, we have identified 7 viable partial solar eclipses in the Terra mission data. We will discuss several results of our study, including comparison of measured SD signal to predicted radiance reduction based on a solar disk radiance model; nominal and outlier behavior as a function of bands, detectors, and mirror-sides; and comparison with other data sets. Our main conclusion from this study is that there is no notable correlation of detector-dependent trend with radiance level for most RSB bands.

S-NPP VIIRS has been on-orbit for more than nine years since it was launched on October 28th, 2011. The VIIRS reflective solar bands (RSBs) are calibrated on-orbit primarily by an onboard solar diffuser (SD). The SD on-orbit degradation is tracked by an onboard SD stability monitor (SDSM). The VIIRS RSBs view the SD through a rotating telescope assembly (RTA). The RTA views the SD from a direction that is quite different from that of the SDSM. It has been shown that the SD degrades non-uniformly with respect to the incident and outgoing directions, especially at the short wavelengths. Thus, the SDSM calibration cannot provide an accurate SD degradation estimation for the view direction of the RTA, resulting in long-term drifts in the calibration coefficients derived from the SD and SDSM calibration. S-NPP VIIRS has been scheduled to view the Moon approximately monthly since its launch. The lunar observations can provide accurate long-term trends for the RSB calibration coefficients since the lunar surface reflectance is quite stable. By comparing the SD and lunar calibration results, we can obtain the SD degradation differences at the view directions of the SDSM and RTA and derive the SD degradation at the view direction of the RTA. Moreover, we can also derive the SD degradation in the short-wave infrared (SWIR) spectral range, wherein the SDSM cannot track the SD degradation. In this paper, we will derive the SD degradation for the view direction of the RTA from the SD, SDSM, and lunar calibrations from the visible to SWIR spectral range. We will also simulate the SD degradation with analytical models and compare their performances.

Computational modeling of spectral and hyperspectral imagery can be performed using radiative flux calculations on highly resolved geometric models. Faceted geometry models are both memory intensive and computationally expensive but allow for a fine-grained approach to radiative modeling. Using high resolution faceted geometry, improved synthetic imagery can be generated from a ray casting sensor model. This paper describes the results of a distributed memory ray tracing architecture for processing high facet count geometry that is capable of modeling radiative flux for highly resolved landscapes. Monte Carlo integration of the radiative transfer equation is coupled with a soil heat transfer model to facilitate solving for temperatures. Ray tracing procedures then use material properties to communicate radiative flux back to a sensor model. Emitted radiation along with mid-wave radiation reflected from neighboring facets and reflected short-wave solar radiation is computed and returned for rays cast from a sensor model. Radiative results of a prototype rainforest have been acquired that demonstrate the modeling capability of the architecture for geometries exceeding 40 million facets. Images of individual spectral components visually validate the legitimacy of the flux simulation. This paper presents an architecture that has been developed with the potential to produce quality synthetic spectral data based on modeling of actual temperature and radiative flux.

In this work we utilize generative adversarial networks (GANs) to synthesize realistic transformations for remote sensing imagery in the multispectral domain. Despite the apparent perceptual realism of the transformed images at a first glance, we show that a deep learning classifier can very easily be trained to differentiate between real and GAN-generated images, likely due to subtle but pervasive artifacts introduced by the GAN during the synthesis process. We also show that a very low-amplitude adversarial attack can easily fool the aforementioned deep learning classifier, although these types of attacks can be partially mitigated via adversarial training. Finally, we explore the features utilized by the classifier to differentiate real images from GAN-generated ones, and how adversarial training causes the classifier to focus on different, lower-frequency features.

We present a fuzzy logic approach allowing the identification of minerals from re ectance spectra acquired by hyperspectral sensors in the VNIR and SWIR ranges. The fuzzy logic system is based on a human reasoning. It compares the positions of the main and secondary absorptions of the unknown spectrum (spectral characteristics estimated beforehand) with those of a reference database (derived from mineralogical knowledge). The proposed solution is first evaluated on laboratory spectra. It is then applied to airborne HySpex and satellite-borne PRISMA images acquired during a dedicated campaign over two quarries in France. This demonstrates the relevance of the method to automatically identify minerals in different mineralogical contexts and in the presence of mixtures.

We present an automatic clustering algorithm for hyperspectral imagery of cultural heritage artifacts: the Selden Map and the Gough Map, both medieval artifacts imaged in the collections at the Bodleian Library of Oxford University. Unlike "traditional" remotely sensed hyperspectral data, these images are of man-made objects using specific materials meant to visually show feature differences and similarities. Consequently, the data are inherently non-Gaussian and potentially very non-linear in the spectral domain. First, we explore the effective graph representations for hyperspectral images, then optimally select the graph modularity to find community structures for a ROI within the scene. By utilizing the eigenvector of the modularity matrix associated with the largest positive eigenvalue for group labeling, we recursively identify multi-level subgroups existing in the graph, producing a variable level of detail cluster-map based on a cluster tree strategy. The generated non-linear decision boundaries are allowed to take any shape with no limits to cluster size. The clustering metric is determined by optimally selecting a high modularity and the largest positive eigenvalue, as well as considering the magnitude of the entry in the leading eigenvector to make each division more accurate and robust. As a result, the optimal number of clusters are found to best characterize the data. Compared to the traditional clustering methods, such as K-means, the graph modularity-based method can extract perceptually important non-local properties of an object, thus yielding semantically more meaningful cluster groups and better discriminating subtle spectral differences between varied pigments. For the Selden Map, we investigate subtle differences in black inks used to denote navigation routes. For the Gough Map, we look at a specific feature under investigation by historians: the castle depicting London. Our results demonstrate the effectiveness of the method as the clustering results explain and match the actual spectra well. This research aims to aid historians in analysis of pigment composition and further facilitate the study in inference of the creation and the evolving timeline of these artifacts.

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.

Experiments or events involving high explosives (HE) can be monitored remotely by infrared (IR) sensors to gather information about the configuration or materials involved in the device. Researchers at the Air Force Institute of Technology (AFIT) developed a phenomenological model for HE fireball spectra in the IR range that allows for parameters to be extracted from Fourier transform infrared (FTIR) data. This model includes parameters tied to physical characteristics of the fireball: temperature, size, soot, and gas concentrations. Previous works have sought to recover these parameters by the fitting of either whole spectra or select wavenumber bands to this phenomenological model. Difficulties arise due to the complex relationships between the parameters to be fit. Uncertainty quantification of the estimated fireball parameters is also problematic since HE experiments do not have any ground truth information on the parameters. It is suggested that artificial neural network (ANN) based approaches may be well suited to this problem, because of their ability to capture complex and highly nonlinear relationships. This work seeks to explore the efficacy of deep artificial neural networks (DNNs) for this problem of parameter recovery from spectra and to also investigate the uncertainty of recovering the fireball parameters from FTIR data. Networks are designed using the hyperparameter optimization tool Hyperopt and trained/tested on artificial data generated using the phenomenological model developed by AFIT. The results of applying the network to the artificial data set are compared to a physics-based band approach that uses a selected number of bands based on their physical properties. Information on the uncertainty of estimating parameters from remotely sensed experimental data is obtained by treating the accuracy of the DNN model on artificial data as an upper bound and by examining the impact of emissivity due to soot on parameter estimation error; the results for artificial data are likely to be optimistic as compared to recovering parameters from experimental data.

Geosynchronous orbit (GEO) weathering induces differential charging of spacecraft surfaces due to simultaneous fluxes of electrons with a wide distribution of energies onto, into, and through spacecraft surface materials. Thus, satellite surfaces can charge thousands of volts with respect to each other whereas entire satellites can charge tens of thousands of volts negative of their surrounding space plasma. The ensuing electric fields can cause local discharges (arcs) from one part of the spacecraft to another, endangering the normal operation of the satellite. Arcing on solar panels can cause reduced optical transmission through solar cell coverglasses which will lead to reduced power production negatively affecting a long-term satellite missions. This work focuses on evaluation of simulated GEO space weather effect, comprised by < 90 keV high-energy electron irradiation, on optical and charge transport properties of two different types of commonly used space solar cells coverglasses, CMX and CMG.

For a given material, a fully characterized bidirectional reflectance distribution function (BRDF) describes how light from any given incident direction reflects into all possible observed directions in space. For simplification, many BRDF measurement and modeling techniques assume isotropic material surface characteristics, focusing primarily on in-plane reflection along individual azimuthal directions. An augmented Complete Angle Scatter Instrument® (CASI®) with a scientific-grade charge-coupled device (CCD) provides the ability to simultaneously capture both in-plane and out-of-plane BRDF data with high spatial resolution, particularly surrounding the specular peak. For any individual CCD frame, each pixel measures the portion of total flux reflected into a unique scatter direction. To properly calculate, analyze, and annotate BRDF readings from raw measurements, each pixel must be mapped to its corresponding scatter direction. This work describes a methodology for mapping pixel location to scatter coordinates based on the geometry of the augmented CASI® system, assuming both the CCD and material surfaces are at. For now, material sample and CCD misalignments are neglected. A broadband metallic laboratory mirror, circularly polished aluminum, and unwrinkled Kapton® samples are then each measured at three incident angles. Measurement results and pixel scatter coordinate mapping are demonstrated for each incident angle, using the beam signature as a proxy for normal incidence. The mirror produces a symmetric specular peak, matching the beam signature, while the polished aluminum and Kapton® produce qualitatively asymmetric specular peaks. Ultimately, this work hopes to foster improvements in BRDF measurement and modeling of materials with anisotropic properties for a range of radiometric simulation, hyperspectral sensing, and scene generation applications.

Previous studies have shown that, for certain data sets, segmentation can help target detection performance for the Matched Filter (MF) algorithm. In this paper, we study the implementation of clustering prior to the Adaptive Cosine Estimator (ACE) calculation and compare our results to the classic non-segmented ACE and Matched Filter algorithms. From our results, we conclude that the proposed algorithm improves Matched Filter results in low false alarm rate conditions, achieving higher accuracy and lower false alarms in target detection; the ACE algorithm results are only marginally affected by segmentation.

Target detection is one of the most important applications utilizing the rich spectral information from hyperspectral imaging systems. Data fusion algorithms applied on hyperspectral datasets address the inherent spatial-spectral resolution tradeoff in these imaging systems by combining spectral information from hyperspectral data with spatial information from hi-res panchromatic or multispectral images (e.g., hi-res RGB). This paper presents the first attempt at using an iterative target implantation technique as a modification to Wald's protocol to assess the performance of data fusion algorithms in target detection tasks. More specifically, this paper looks at how the sharpening process localizes and discriminates the subpixel target from its background, and characterizes an image-wide detectability of any single subpixel target independent of location in the image. We used NNDi use as our pansharpener to perform HRPAN+LRHSI data fusion and the adaptive coherence estimator (ACE) as our target detector. Results show that our methodology is effective at assessing (1) how the sharpening process enhances target-background separability within any 5x5 window anywhere on the image and (2) how the sharpening process enhances the detectability of a single subpixel target over the entire hyperspectral image.

Spectral mixture analysis of hyperspectral imagery allows for detection, classification and quantification of targets present in the imaged scene. It can be difficult to characterize the performance of spectral unmixing techniques at detection, classification and quantification of field data at the subpixel level due to limited ground truth, especially for mixed pixels. Fortunately, the SpecTIR Hyperspectral Airborne Experiment (SHARE) 2012 contains a set of targets specifically designed to test spectral unmixing algorithms. In this paper we explore the performance of an unconstrained and three constrained least squares methods for supervised spectral unmixing. Each of the three methods provides an estimate of the abundance of known targets which can be used for detection, classification and quantification. A detailed evaluation of these spectral unmixing techniques on the SHARE 2012 hyperspectral data is used to demonstrate the performance of each method at supervised target detection, classification and quantification.

Segmentation is useful during sub-pixel target detection in hyperspectral data. Previous works have showed that improving performance in sub-pixel target detection can be achieved by making better estimates of the covariance matrix by using segmentation. One of the challenges mentioned was that pixel assignment has been influenced by the target, and therefore a reassignment after segmentation was needed. We examined and compared several methods to deal with this challenge before the segmentation process, as well as to check if this was essential for our algorithm’s success. Using simulations and several analytical tools we analyzed the matched-filter algorithm, both with and without segmentation, and compare performances of the receiver operating characteristic curves. We found there is no need to perform reassignment after segmentation; segmentation is effective even with the presence of the target in the examined pixel.

Short wave infrared (SWIR) is the reflected radiance in the 1-2.5 μm range. Mid wave infrared (MWIR) is the radiated radiance in the 3-5 μm range. This paper proposes a paint detection method using two infrared bands with different characteristics. Object detection is one of the issues in hyperspectral image (HSI). We use one dimensional convolution neural network (1D-CNN) and guided gradient-weighted class activation mapping (Guided Grad-CAM) for band selection. We make a 1D-CNN architecture and select bands using Guided Grad- CAM from well-trained 1D-CNN. Finally, paint is detected using selected bands. We use datasets included short wave infrared band (SWIR) and mid wave infrared band (MWIR).

Aflatoxins are a family of carcinogenic toxins produced by Aspergillus flavus and Aspergillus parasiticus fungi and occur on agricultural crops such as maize or corn, peanuts, cottonseed, and tree nuts. Even at low concentrations (< 20 ppb), it affects the liver, kidneys, heart, respiratory, nervous, endocrine systems and growth in infants and children. The economic impact of aflatoxin-associated is estimated to be on the order $47M of losses per year in food crops and $225 million per year in feed crops. Current methods for detection include chromatographic methods, spectroscopy and immunochemistry. Although very sensitive, these are largely cumbersome, require extensive sample preparation, skilled operators and in the case of chromatography, very expensive equipment. Hyperspectral imaging offers a sensitive, convenient, compact, reliable, relatively economical and simple to operate solution requiring little or no sample preparation that can be used in the laboratory or the field.

Assessment of panchromatic sharpening algorithms typically starts with a high resolution Hyperspectral Image (HSI), which is then spatially degraded so that after the sharpening process, the result can be compared to the original and analyzed for accuracy. This leads to questions about quantitative assessments based solely on simulated low resolution data. To address this problem, a multi-resolution hyperspectral data set was collected by researchers from the Rochester Institute of Technology (RIT) in Henrietta, NY on July 24th 2020. Imagery of 48 felt targets, ranging in size from 5 cm to 30 cm and in six different colors, was collected using RIT's MX-1 UAS imaging system, which is designed to collect 272 spectral bands from approximately 400nm to 1000nm. Three flights were performed and images of the target scene were collected at flight altitudes of 30m, 60m, and 120m. The resulting imagery possesses ratios of 2:1 and 4:1 spatial resolution relative to the lowest altitude flight. The goal of this imaging campaign was to create a data set that will be used to test hyperspectral pansharpening algorithms currently under development at RIT. The radiometric accuracy of sharpening algorithms can be better ascertained through quantitative analysis of their results after being applied to this non-simulated multiresolution hyperspectral data set. This presentation will summarize the process of planning, creating, collecting and packaging the dataset which will be provided to RIT researchers and will also available for download online through RIT.

In this paper we describe a web-based application we have developed that allows visualization of the multi-spectral reflectance properties of three-dimensional surfaces. To develop this application, we employ a 3D computer graphics framework that enables us to combine spectral data with other object features such as topography and reflectance properties. The cross-platform application runs in any modern web browser, has an interactive interface, and does not require the installation or maintenance of any custom or proprietary software packages. The application is configured as a standard HTML/JavaScript client-server system, which provides flexibility for working with data in local or remote setups.

Detection, identification, and quantification of greenhouse gases is essential to ensure compliance with regulatory guidelines and mitigate damage associated with anthropogenic climate change. Passive infrared hyperspectral imaging technology is among the solutions that can detect, identify and quantify multiple greenhouse gases simultaneously. The Telops Hyper-Cam Airborne Platform is an established system for aerial thermal infrared hyperspectral measurements for gas survey applications. In support of the Hypercam, Telops is developing a suite of hyperspectral imaging data processing algorithms that allow for gas detection, identification, and quantification in real-time. In the Fall of 2020, the Hyper-Cam-LW Airborne platform was flown above a validated SF6 gas release system to collect hyperspectral data for gas quantification analysis. This measurement campaign was performed to document performance of the Hyper-Cam gas quantification capabilities against known quantities of released gas. This talk introduces the principles behind the gas detection, identification, and quantification algorithms and presents the motivations and results from the Fall 2020 measurement campaign.

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.

Cesium lead halide perovskites are characterized by unique dielectric response properties that are significant for potential applications. Modeling the dielectric response of these materials with respect to underlying physical processes is therefore useful. A case study analysis is presented of the dielectric response of cesium lead halide perovskites, using computational experiments and theory, helping to formulate quantitative dielectric response functions. In particular, the electronic structure and the dielectric function can be calculated using computational methods based on density function theory (DFT). Our analysis uses the Vienna Ab-initio Simulation Package (VASP), where the dielectric function is calculated by solving the Bethe-Salpeter equation on top of electronic structure calculations within the GW0 method.

Hyperspectral imaging systems are used in industrial sorting solutions to differentiate materials by subtle spectral features (spectral fingerprint) which are not resolvable by the human eye. Prominent applications are the detection of contaminants in food or the separation of different types of plastic. Hyperspectral line scan systems are suitable for many high throughput applications. A line across the sample, perpendicular to the direction of the relative movement, is projected into an imaging spectrograph. The spectral information for each pixel along this line is projected along the second axis of the two-dimensional detector chip. By spatial scanning of the sample the spectral data cube gets recorded. This paper presents the development of a laboratory hyperspectral imaging system capable of acquiring spectral data in the range from 400nm to 1700 nm. Therefore, Specim hyperspectral cameras FX10 and FX17 are used. The laboratory setup mimics an industrial sorting machine and will be used for application development. State of the art machine learning algorithms are used for data classification. Selected sorting applications are presented.

The increasing number of successfully deployed space missions have resulted in an increased density of manmade objects positioned in orbital domains near Earth. With this steady accumulation of objects in space, it is becoming more imperative to characterize spacecraft materials, which may ultimately be contributors to the orbital debris population. In order to ascertain the potential damage from orbital debris, a laboratory hypervelocity impact test was conducted using a 56-kg modern spacecraft representative satellite (DebriSat) to simulate a catastrophic fragmentation event in low Earth orbit. In an effort to identify unique, material-specific spectroscopic markers, a select number of the spacecraft materials used to construct DebriSat were analyzed using reflectance spectroscopy as a characterization technique for assessment on material type according to optical features. Spectral measurements of DebriSat materials analyzed prior to the laboratory impact are presented in this paper. These data provide a spectral characterization baseline for modern-day spacecraft materials in their pristine conditions and are compared to each other to distinguish spectra of materials belonging to different classifications with an effort of grouping them using color index. The ongoing efforts to classify materials utilizing their reflectance spectroscopic fingerprint are discussed in this study.

As West Nile Virus (WNV) and St. Louis Encephalitis (SLE) become more prevalent across North America, there is an increased risk of fatal neuroinvasive cases. In order for public health officials to prepare for these cases and potentially intervene, the ability to forecast mosquito borne disease outbreaks is paramount. In practice, however, such vector borne diseases are notoriously difficult to predict due to their seemingly sporadic spatial and temporal outbreak patterns. Recent research has demonstrated that mosquito abundance is causally related to WNV/SLE prevalence, providing a practical starting point for developing mosquito-borne disease forecasting systems. When focusing on building mosquito population models, understanding the reproduction environment of Culex mosquitos (WNV and SLE's primary vectors) is key: they rely on warmth, water, and vegetation to reproduce. Previous work has shown that global-coverage multispectral imagery (MSI) (i.e., Landsat 8, Sentinel- 2) is a valuable resource for characterizing vegetation health as a predictor of mosquito population, but it is limited in that it may not provide the spatial resolution necessary to distinguish between, e.g., a well-fertilized lawn (poor Culex habitat) and a stand of trees (good Culex habitat). The backscatter information collected by synthetic aperture radar (SAR) imagery provides opportunity to distinguish between broader categories of vegetation type, potentially helping to fill this gap. This research uses publicly available global-coverage MSI and SAR imagery (Landsat 8, Sentinel-2, and Sentinel-1) to explore if vegetation type, in tandem with vegetation health, improves our ability to forecast mosquito populations. Vegetation characterization is done over the Greater Toronto Area from 2014 to 2017, and we derive weekly time series from MSI, spectral indices, and SAR for this time period. We then quantify the strength of vegetation health and type as a predictor of Culex abundance.

Hyperspectral imaging systems act as imaging spectrometers, acquiring dozens or hundreds of equally spaced spectral channels, thus leading to high complexity setups, low acquisition speed, and a large amount of data. In order to optimize the number of acquisition channels, and to mitigate these problems in shortwave infrared (SWIR) hyperspectral imaging systems, one must extend to the SWIR range the analysis and characterization methods that are available, in the literature, for the visible spectrum. To that end, this work focuses on the SWIR surface spectral reflectance (SSR) of possible objects that may be present in the scene, by analyzing an empirical SSR library that includes the SWIR range, as the ECOSTRESS spectral library. To the best of our knowledge, this is the first report of SWIR SSR data analysis in this library. The main goal of this paper is to investigate the approximation of data samples in this library by two linear models, namely truncated Fourier Series and principal components, both with less than a dozen basis vectors. This corresponds to significant dimension reduction in comparison to the number of acquisition channels, which lies in the several hundreds in the ECOSTRESS library. To validate the analysis and assess the quality of the reconstructed spectra, root mean squared error and goodness-of-fit coefficient(GFC) metrics are applied. An `accurate' to `excellent' fit with GFC median ranging between 0.995 and 0.9999 is obtained when reconstructing the signals with three to five principal components, and greater than 0.995 with three to five Fourier series terms.

This paper explores data summarization methods to find reduced size representations of hyperspectral data. Operating on the small size summary reduces computational requirement further up in the image processing chain. We look into how to construct such summarizations and their application in endmember extraction for unmixing. We compare methods based on random sampling and methods based on superpixel representations. We show that summaries based on superpixel analysis better summarize the characteristics of the image compared to random sampling. Results with real images are presented.