Reflectance obtained from first principles atmospheric correction algorithms can be improved using spectral polishing. We describe a new polishing method that leverages a library of spectrally smooth materials to virtually eliminate atmospheric residuals.

In this tutorial overview, we examine atmospheric compensation of hyperspectral data in the visible and near-infrared (VNIR)-short-wave infrared (SWIR) region.

This study examines how hyperspectral rare target detection performance is affected by the method of atmospheric compensation used to convert the data to reflectance units. Rare and subpixel target detection algorithms employ contrast enhancement methods to suppress signatures from the background materials. Therefore, when evaluating atmospheric compensation methods, it is important to consider their accuracy in contrast-enhanced space. In particular, a key requirement for good detection is the suppression of atmospheric band residuals in the reflectance spectra, making them as smooth as possible. This explains the success of the empirical Quick Atmospheric Correction (QUAC) algorithm and the importance of supplementing first principles methods with spectral polishing. We illustrate these findings using two data sets acquired by the Rochester Institute of Technology (RIT), three different whitening-based detection algorithms, and three different atmospheric compensation algorithms, QUAC, FLAASH and ATCOR.

A lightweight, high-resolution spectral imager was developed for aerial surveys of vegetative traits using small drones. The sensor has the potential to rapidly map vegetative type, health, and function over large areas, and revisit the same area to measure seasonal changes in vegetative mass, water uptake, and other traits. It will be invaluable for the growing field of precision agriculture and for ecosystem research. The key benefit, compared to other existing hyperspectral sensors, is the extended spectral range, from the visible into the short-wave infrared, which enables measurement of vegetative features that cannot be accurately measured with existing visible sensors. In Phase II, we developed a small, low-weight, hyperspectral imaging system for UASs based on a high-performance solid-core spectrograph and robust processing algorithms that produce calibrated vegetation trait products. The key innovation was the design and fabrication of solid-block spectrometer machined from a transparent crystal that has an extended spectral range from the visible to the short-wave-infrared (0.38-2.45 microns). It is coupled with a wide-field of view telescope and extended-range camera to produce hyperspectral image (2D images in 400 different colors) with extremely high spatial and spectral resolution. The spectrograph is packaged with ancillary electronics, GPS, and inertial navigation, into a small, 5-pound package that attaches to a standard multicopter drone. SSI is teamed with DOE researchers to develop the science behind measuring vegetative traits from the air, concentrating on how spectral traits correlate from the leaf scale to the landscape scale. In Phase II we have demonstrated the flight package and mapping techniques using a surrogate spectrometer with a smaller spectral range. In a follow-on Phase IIA program, we will be flying the extended-range sensor in coordinated field tests to correlate the landscape-scale hyperspectral measurements with ground measurements. The new extended-range sensor will be integrated into a commercial product line aimed at the precision agriculture and mineral exploration industries. It represents a breakthrough in size, weight, and performance that will enable routine application in the growing, but cost-sensitive field of precision agriculture. It will also be sold for ecological research. Large-scale production of the sensor will enable systematic, ecosystem-scale surveys of plant type, plant-health, carbon-uptake, and vegetative mass.

Within the last few years, several commercial long-wave infrared (LWIR) hyperspectral imaging (HSI) systems have been developed for remote sensing of the ground from aircraft. While much less expensive and more practical to operate than sensors such as SEBASS and MAKO, which have been developed primarily for research and Government use, the commercial systems have poorer signal-to-noise and/or spectral resolution. We investigate the utility of three commercial systems-the Telops Hyper-Cam, SPECIM AisaOWL, and ITRES TASI-600-for quantitative retrieval of surface temperature and emissivity spectra. Atmospheric retrieval, correction and temperature-emissivity separation are performed on example data from these sensors using FLAASH-IR, a first-principles algorithm that incorporates radiation transport calculations and atmosphere models from MODTRAN. The results from the commercial sensors are noisy compared with SEBASS but otherwise appear to be reasonable. Applying a noise suppression algorithm to the radiance data yields better temperature retrievals and much cleaner emissivity spectra, with minimal loss of information, and should benefit scene classification applications.

We describe a new algorithm, QUAC-IR (QUick Atmospheric Correction in the InfraRed), for automated, fast, atmospheric correction of LWIR (Long Wavelength InfraRed) hyperspectral imagery (HSI) and multi-spectral imagery (MSI) in the ~7-14 mm spectral region. QUAC-IR is an in-scene based algorithm, similar to the widely used ISAC (In- Scene Atmospheric Correction) algorithm. It improves upon the ISAC approach in several key ways, including providing absolute, versus relative, sensor-to-ground transmittances and radiances, as well as an estimate of the atmospheric downwelling sky radiance. The latter is important for retrieving emissivity from a reflective (i.e., non-blackbody) pixel. The key aspect of QUAC-IR is that it explicitly searches for blackbody pixels using an efficient approach involving a small number of spectral channels in which the atmospheric radiative transfer is dominated by the water continuum. This allows for fast and simplified Beer's Law (i.e., exponential) scaling of the path transmittance and radiance based on a compact library of pre-computed reference values. We apply QUAC-IR to well-calibrated data from the SEABASS1 and MAKO2 HSI sensors. The results are compared to those from a first-principles physics-based atmospheric code, FLAASH-IR.

Extremely thick haze caused by air pollution is observed in many satellite images of the earth, and in particular over eastern China. Standard image display software typically provides satisfactory visualization of the ground through automated or user-driven scaling to enhance contrast; however, it does not perform well with these highly polluted scenes, where the haze is spatially non-uniform. Furthermore, estimation of surface reflectance using standard atmospheric correction software is highly problematic under these conditions due to very low visible transmission of the haze coupled with lack of knowledge of its optical properties, which may not conform to the haze or aerosol models in the software. In this paper we show that a version of the empirical Quick Atmospheric Correction (QUAC) algorithm, adapted for spatially dependent scattering, produces visually satisfying imagery of the entire ground in multispectral satellite scenes containing thick haze, and that the output reflectance spectra appear to be realistic enough for performing basic surface classification. The QUAC algorithm is applicable to multispectral and hyperspectral imagery with any number of wavelength bands, including true color (RGB) imagery, and does not require radiometrically calibrated data.

Composite materials are widely used in aircraft to reduce manufacturing costs, improve structural performance, and boost fuel efficiency. However, safety concerns arise because of the susceptibility of these composites to inadequate adhesive bond quality, including so-called “kissing bonds” which may occur because of initial fabrication or service-related issues. Such weakened bonds may have occurred due to contamination at the bondline surface or in the adhesive, imperfect adhesive thermal cure, or inadequate mixing of adhesive constituents. Because of the widespread and rapidly growing use of composite materials in military aircraft, there is a need for an easily used detection approach to routinely monitor the health of composite materials, both in the factory and in deployed aircraft. To address this need, we have developed a new optical/acoustic approach to test bond quality in composite materials. Lamb wave vibrations are induced in the sample with a high frequency transducer and are visualized using motion-contrast laser speckle imaging with a fast-framing commercial camera. In measurements conducted on test panel sandwiches prepared from SGP370-8H 8552 fiber/epoxy plates and epoxy adhesive, the processed video imagery reveals information on the presence and location of bond flaws in the adhesive layer.

Classification of objects, materials or terrain in hyperspectral imagery requires the definition of an appropriate measure of spectral similarity, typically expressed in terms of spectral reflectance. For many objects, absolute reflectance varies due to bidirectional reflectance distribution function (BRDF) effects or uneven illumination. Here, an appropriate similarity measure is spectral angle; however, spectral distance can be used instead if the data are amplitude normalized. Further improvement in classification can be obtained with a regularized whitening step that normalizes the data spread along leading principal component coordinates but limits the spread within trailing principal components. We demonstrate this approach using a Support Vector Machine (SVM) classifier to identify vehicles across hyperspectral images in a time sequence. The normalization and whitening preconditioning steps lead to similar classification performance using spectral distance to the target mean, a much simpler and faster method than the SVM.

In this paper, the detection of point target in a target rich environment is considered. The standard matched filter may be problematic if the estimate of the background is contaminated by neighboring pixels. A second solution, the Orthogonal Space Projection (OSP) implementation of the Generalized Likelihood Ratio Test may be faulty due to the over-definition of the background with the use of too many endmembers. Throughout this paper, a modified version of the OSP test (MOSP) is developed; the MOSP performs well in a point target detection test.

Multispectral and hyperspectral imaging can facilitate vehicle tracking across a series of images by gathering spectral information that distinguishes the vehicle of interest from confusers. Developing effective algorithms for utilizing this information requires an understanding of the sources and nature of both the common and unique components in vehicle spectra, as well as the variations associated with lighting, view angle, and part of the vehicle being observed. In this study, focusing on the VNIR-SWIR spectral region, we analyze hyperspectral data from a recent field experiment at the Rochester Institute of Technology. We describe the spectra of painted vehicle surfaces in general terms, and demonstrate effective classification of automobiles based on spectra from upward facing surfaces (the roof, hood or trunk) using a method that combines the Support Vector Machine with data pre-conditioning.

Algorithms for retrieval of surface reflectance, emissivity or temperature from a spectral image almost always assume uniform illumination across the scene and horizontal surfaces with Lambertian reflectance. When these algorithms are used to process real 3-D scenes, the retrieved "apparent" values contain the strong, spatially dependent variations in illumination as well as surface bidirectional reflectance distribution function (BRDF) effects. This is especially problematic with horizontal or near-horizontal viewing, where many observed surfaces are vertical, and where horizontal surfaces can show strong specularity. The goals of this study are to characterize long-wavelength infrared (LWIR) signature variability in a HSI 3-D scene and develop practical methods for estimating the true surface values. We take advantage of synthetic near-horizontal imagery generated with the high-fidelity MultiService Electro-optic Signature (MuSES) model, and compare retrievals of temperature and directional-hemispherical reflectance using standard sky downwelling illumination and MuSES-based non-uniform environmental illumination.

Hyperspectral imaging (HSI) sensors have the ability to detect and identify objects within a scene based on the distinct attributes of their surface spectral signatures. Many targets of interest, such as vehicles, represent a complex arrangement of specular (non-Lambertian) materials with curved and flat surfaces oriented at varying view factors. This complexity, combined with possible changing atmospheric/illumination conditions and viewing geometries, can produce significant variations in the observed signatures from measurement to measurement, making detection and/or reacquisition challenging. This paper focuses on the characterization of visible-near infrared-short wave infrared (VNIR-SWIR) spectra for detection, identification and tracking of vehicles. Signature variations are predicted using a novel image simulation tool to calculate spectral images of complex 3D objects from a spectral material description such as the modified Beard-Maxwell BRDF model, a wireframe shape model, and a directional model of the illumination. We compare the simulations with recent VNIR-SWIR hyperspectral imagery of vehicles and panels collected at the Rochester Institute of Technology during an Autumn 2015 measurement campaign. Variations in both the simulated and measured spectra arise mainly from differences in the relative glint contribution. Implications of these variations on vehicle detection and identification are briefly discussed.

Atmospheric compensation algorithms for LWIR hyperspectral imagery typically require the presence of blackbody surfaces, resulting in failure with many desert scenes. The FLAASH-IR algorithm is successfully applied to challenging SEBASS and HyTES desert images.

In recent years long-wavelength infrared (LWIR) hyperspectral imagery has significantly improved in quality and become much more widely available, sparking interest in a variety of applications involving remote sensing of surface composition. This in turn has motivated the development and study of LWIR-focused algorithms for atmospheric retrieval, temperature-emissivity separation (TES) and material detection and identification. In this paper we evaluate some LWIR algorithms for atmospheric retrieval, TES, endmember-finding and rare material detection for their utility in characterizing mineral composition in SEBASS hyperspectral imagery taken near Cuprite, NV. Atmospheric correction results using the In-Scene Atmospheric Correction (ISAC) method are compared with those from the first-principles, MODTRAN©-based FLAASH-IR method. Covariance-whitened endmember-finding methods are observed to be sensitive to image artifacts. However, with clean data and all-natural terrain they can automatically locate and distinguish many minor mineral components, with especially high sensitivity to varieties of calcite. Not surprisingly, the major scene materials, including silicates, are best located using unwhitened techniques. Minerals that we identified in the data include calcite, quartz, alunite and (tentatively) kaolinite.

This talk summarizes the current state of the art in atmospheric effects modeling and compensation for HSI, distinguishing straightforward from difficult cases, assessing approximate RT schemes, and highlighting current directions and unsolved problems.

The quick atmospheric correction (QUAC) algorithm is a relatively fast and robust atmospheric compensation algorithm for hyperspectral image processing utilizing in scene information. An adjustment of some key parameters in QUAC is made leading to improved results for coastal scenes. In general the QUAC results compare well with two first principles radiative transfer (RT) model based algorithms. Some suggestions for future work are made including automating the setting of key QUAC parameters and accounting for the coastal zone aerosols more accurately in the RT algorithms.

This study examines the influence of non-Lambertian reflectance effects on the detection of subpixel vehicles in hyperspectral imagery. Object-level BRDF spectral signatures for an olive green sedan were simulated using a fast, radiometrically accurate 3D rendering model, and the signatures were embedded as subpixel targets of low fractional fill into the HyMap hyperspectral image provided in the Rochester Institute of Technology Blind Test dataset. Detection algorithms based on the ACE detector were run on the scene. The results demonstrate a significant improvement in detection performance when including the target's BRDF variation in the detection scheme through either a subspace ACE or a Bayesian selection (multiple ACE detector) method.

Processing long-wave infrared (LWIR) hyperspectral imagery to surface spectral emissivity or reflectance units via atmospheric compensation and temperature-emissivity separation (TES) affords the opportunity to remotely classify and identify surface materials with minimal interference from atmospheric effects. This paper describes an automated atmospheric compensation and TES method, called FLAASH-IR (Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes Infrared), and its application to airborne imagery taken with the Telops Inc. Hyper-Cam interferometric hyperspectral imager. The results demonstrate good suppression of the atmospheric features due to water vapor and ozone, resulting in quantitative surface spectra, even with highly reflective (low emissivity) objects such as bare metal.