Search

The emergence of machine learning into scientific fields has created opportunities for novel and powerful image processing techniques. Algorithms that can perform complex tasks without a “man-in-the-loop” or explicit instructions are invaluable artificial intelligence tools. These algorithms typically require a large set of training data on which to base statistical predictions. In the case of electro-optical infrared (EO/IR) remote sensing, algorithm designers often seek a substantial library of images comprising many weather conditions, times of day, sensor resolutions, etc. These images may be synthetic (predicted) or measured, but should encompass a large variety of targets imaged from a variety of vantage points against numerous backgrounds. Acquiring such a large set of measured imagery with sufficient variation can be difficult, requiring numerous field campaigns. Alternatively, accurate prediction of target signatures in cluttered outdoor scenes may be a viable option. In this work, sensor imagery is generated using CoTherm, a co-simulation tool which operates MuSES (an EO/IR simulation code) in an automated fashion to create a large library of synthetic images. The relevant MuSES inputs – which might include environmental factors, global location, date and time, vehicle engine state, human clothing and activity level, or sensor waveband – can be manipulated by a CoTherm workflow process. The output of this process is a large library of MuSES-generated EO/IR sensor radiance images suitable for algorithm development. If desired, synthetic target pixels can be inserted into measured background images for added realism.

Optical propagation is strongly influenced b y t he n umber concentration, size distribution, thermodynamic phase, and spatial distribution of particles in atmospheric clouds. These properties have been investigated in the field using an airborne digital holographic instrument. A laboratory facility has also been developed, in which optical propagation is being investigated in steady-state turbulent-cloud conditions.

The ability to accurately predict electro-optical signatures for high-value targets in an outdoor scene is a tremendous asset for defense agencies. In the thermal infrared wavebands, physical temperature is the primary contributor to imager-detected radiance. Consequently, enhancements to the fidelity of thermal predictions are desirable, and convection estimates are often the most significant source of uncertainty. One traditional method employed by MuSES applies a single global convection coefficient across the entire scene, and assumes that all exposed surfaces are in contact with the ambient air temperature. This results in a rapid prediction of convection coefficients for a large scene that changes dynamically with wind speed but lacks localized detail concerning how each target surface can experience a different convection coefficient and local air temperature. In practice, wind speed and directions typically change frequently, which coupled with the thermal mass of most targets reduces the negative impacts this approach can have on thermal predictions; numerous validations of MuSES predictions bear this out. Computational fluid dynamics (CFD) simulations provide additional spatial fidelity in the calculation of localized convection coefficients and air temperatures but only at great computational cost. A fully transient outdoor scene simulation would be nearly impossible. Boundary layers near surfaces must be resolved with a fine mesh, creating a numerical problem difficult to solve for large spatial scales when spanning large periods of simulated time. In this paper, a novel thermal fluid flow solver is presented. This proprietary flow solver models fluid flow at the spatial resolution and accuracy needed for convective heat transfer at the scale viewed by EO/IR (Electro-optical/Infrared) sensors, avoiding the burdens associated with conventional CFD codes. Many of the same Navier-Stokes equations are solved, albeit in simpler form. ThermoAnalytics-developed correlations directly calculate convection coefficients based on local bulk flow conditions of temperature, velocity, and pressure. The resulting accuracy is a large step beyond using a constant convection correlation universally across the scene. Intelligent simplification of the flow equations provides robust efficiency, requiring minimal effort and expertise compared to CFD codes.

The ability to predict spectral electro-optical (EO) signatures for various targets against realistic, cluttered backgrounds is paramount for rigorous signature evaluation. Knowledge of background and target signatures, including plumes, is essential for a variety of scientific and defense-related applications including contrast analysis, camouflage development, automatic target recognition (ATR) algorithm development and scene material classification. The capability to simulate any desired mission scenario with forecast or historical weather is a tremendous asset for defense agencies, serving as a complement to (or substitute for) target and background signature measurement campaigns. In this paper, a systematic process for the physical temperature and visible-through-infrared radiance prediction of several diverse targets in a cluttered natural environment scene is presented. The ability of a virtual airborne sensor platform to detect and differentiate targets from a cluttered background, from a variety of sensor perspectives and across numerous wavelengths in differing atmospheric conditions, is considered. The process described utilizes the thermal and radiance simulation software MuSES and provides a repeatable, accurate approach for analyzing wavelength-dependent background and target (including plume) signatures in multiple band-integrated wavebands (multispectral) or hyperspectrally. The engineering workflow required to combine 3D geometric descriptions, thermal material properties, natural weather boundary conditions, all modes of heat transfer and spectral surface properties is summarized. This procedure includes geometric scene creation, material and optical property attribution, and transient physical temperature prediction. Radiance renderings, based on ray-tracing and the Sandford-Robertson BRDF model, are coupled with MODTRAN for the inclusion of atmospheric effects. This virtual hyperspectral/multispectral radiance prediction methodology has been extensively validated and provides a flexible process for signature evaluation and algorithm development.

Radiative transfer through particle-laden media such as clouds can be impacted by variations in particle spatial distributions. Due to mixing and inertial effects of droplets suspended in the almost always turbulent atmosphere, cloud particles are often spatially-correlated. The correlations result in clusters and voids within the droplet field that, even when smaller than the photon mean free path, can lead to deviations from the exponential extinction law. Prior work has numerically investigated these departures from exponential attenuation in absorptive media; this work extends those results for a scattering medium. The problem is explored with a Monte Carlo Ray Tracing (MCRT) program capable of tracking light attenuation through both perfectly random (uncorrelated) and spatially correlated collections of scatterers and/or absorbers. The MCRT program is favorably compared to two-stream flux equations, and numerical exploration of the pure-absorption case is used to determine the sampling statistics necessary to characterize radiative transmission within the numerical simulation. Light transmission through fields of spatially-correlated, non-absorbing, scattering particles is explored. Particles are distributed following a Matern Point Process, which allows cluster strength and size, as well as the usual variables of particle scattering cross section and number density to be varied. The results show that the degree of non-exponential attenuation is determined by the magnitude and shape of the radial distribution function, which describes correlations in discrete (non-continuous) particle distributions. Parametric studies revealed that the number of clusters and cluster radius, factors in the Matern radial distribution function, impact direct, diffuse and backward radiative transfer. The Matern RDF is shown to be consistent with a previous “cloudlet” approach, providing a bridge between the analytical cloudlet model and continuous correlation function approaches.

The ability to predict electro-optical (EO) signatures of diverse targets against cluttered backgrounds is paramount for signature evaluation and/or management. Knowledge of target and background signatures is essential for a variety of defense-related applications. While there is no substitute for measured target and background signatures to determine contrast and detection probability, the capability to simulate any mission scenario with desired environmental conditions is a tremendous asset for defense agencies. In this paper, a systematic process for the thermal and visible-through-infrared simulation of camouflaged human dismounts in cluttered outdoor environments is presented. This process, utilizing the thermal and EO/IR radiance simulation tool TAIThermIR (and MuSES), provides a repeatable and accurate approach for analyzing contrast, signature and detectability of humans in multiple wavebands. The engineering workflow required to combine natural weather boundary conditions and the human thermoregulatory module developed by ThermoAnalytics is summarized. The procedure includes human geometry creation, human segmental physiology description and transient physical temperature prediction using environmental boundary conditions and active thermoregulation. Radiance renderings, which use Sandford-Robertson BRDF optical surface property descriptions and are coupled with MODTRAN for the calculation of atmospheric effects, are demonstrated. Sensor effects such as optical blurring and photon noise can be optionally included, increasing the accuracy of detection probability outputs that accompany each rendering. This virtual evaluation procedure has been extensively validated and provides a flexible evaluation process that minimizes the difficulties inherent in human-subject field testing. Defense applications such as detection probability assessment, camouflage pattern evaluation, conspicuity tests and automatic target recognition are discussed.

Earth’s atmosphere can significantly impact the propagation of electromagnetic radiation, degrading the performance of imaging systems. Deleterious effects of the atmosphere include turbulence, absorption and scattering by particulates. Turbulence leads to blurring, while absorption attenuates the energy that reaches imaging sensors. The optical properties of aerosols and clouds also impact radiation propagation via scattering, resulting in decorrelation from unscattered light. Models have been proposed for calculating a point spread function (PSF) for aerosol scattering, providing a method for simulating the contrast and spatial detail expected when imaging through atmospheres with significant aerosol optical depth. However, these synthetic images and their predicating theory would benefit from comparison with measurements in a controlled environment. Recently, Michigan Technological University (MTU) has designed a novel laboratory cloud chamber. This multiphase, turbulent “Pi Chamber” is capable of pressures down to 100 hPa and temperatures from -55 to +55°C. Additionally, humidity and aerosol concentrations are controllable. These boundary conditions can be combined to form and sustain clouds in an instrumented laboratory setting for measuring the impact of clouds on radiation propagation. This paper describes an experiment to generate mixing and expansion clouds in supersaturated conditions with salt aerosols, and an example of measured imagery viewed through the generated cloud is shown. Aerosol and cloud droplet distributions measured during the experiment are used to predict scattering PSF and MTF curves, and a methodology for validating existing theory is detailed. Measured atmospheric inputs will be used to simulate aerosol-induced image degradation for comparison with measured imagery taken through actual cloud conditions. The aerosol MTF will be experimentally calculated and compared to theoretical expressions. The key result of this study is the proposal of a closure experiment for verification of theoretical aerosol effects using actual clouds in a controlled laboratory setting.

Optical characterization of space objects generally involves the observation of an object of interest as it passes over a ground-based imaging asset. Owing to the presence of turbulence, adaptive optics, post-processing, or a combination thereof, is used to enhance imagery. Specular glints, often present in such imagery, can introduce severe artifacts after post-processing methods are applied. Removing these artifacts generally involves excluding frames that include glints. With fewer frames, image reconstruction quality is reduced and in some cases entire collections are discarded. We describe a method where Robust Alignment by Sparse Low-rank decomposition (RASL) is used to identify and exclude glints from sets of turbulence-corrupted imagery. A number of simulated image sets including glints were generated and the RASL algorithm applied. We show that RASL is capable of identifying glints in these images. Once identified, glints are removed or masked. Post-processing using corrected imagery results in reconstructions free from severe artifacts.

Deleterious effects of the atmosphere on remotely acquired images includes absorption and scattering of light by aerosol particulates, which not only attenuates the signal but can potentially cause blurring due to forward-scattered light accepted by the imaging system. Proposed aerosol scattering models (e.g., Ishimaru) provide a method for simulating the contrast and spatial detail expected when imaging through atmospheres with significant aerosol optical depth. This work explores closure between modulation transfer functions (MTFs) obtained from directly measured images and MTFs calculated from theory using measured cloud properties. The closure experiments are performed in a laboratory cloud chamber in which cloud droplet number density and size distribution are directly measured. Images of a binary knife-edge target were taken with an optical detector on the other side of a water cloud generated through reduction of pressure in the humidified chamber. The key results of this closure experiment are: the theoretical expression for the aerosol MTF is likely overly simplistic and does not account for broad particle size distributions. The significance of optical blurring from light scattering by aerosol particles depends sensitively on the properties of both the particles and the imaging system.

Accurate infrared signature prediction of targets, such as humans or ground vehicles, depends primarily on the realistic prediction of physical temperatures. Thermal model development typically requires a geometric description of the target (i.e., a 3D surface mesh) along with material properties for characterizing the thermal response to simulated weather conditions. Once an accurate thermal solution has been obtained, signature predictions for an EO/IR spectral waveband can be generated. The image rendering algorithm should consider the radiative emissions, diffuse/specular reflections, and atmospheric effects to depict how an object in a natural scene would be perceived by an EO/IR sensor. The EO/IR rendering process within MuSES, developed by ThermoAnalytics, can be used to create a synthetic radiance image that predicts the energy detected by a specific sensor just prior to passing through its optics. For additional realism, blurring due to lens diffraction and noise due to variations in photon detection can also be included, via specification of sensor characteristics. Additionally, probability of detection can be obtained via the Targeting Task Performance (TTP) metric, making it possible to predict a target’s at-range detectability to a particular threat sensor. In this paper, we will investigate the at-range contrast and detectability of some example targets and examine the effect of various techniques such as sub-pixel sampling and target pixel thresholding.

In principle, infrared (IR) imagery of exposed skin and clothing can provide a valuable source of data for thermo-physiological studies. In practice, and despite the fall in the cost of IR cameras, infrared imagery is not universally collected during human subject testing. One reason for this may be the relatively poor accuracy of IR cameras (typically ±2 °C). The repeatability of the measurements taken with a particular IR camera (both spatially and temporally) is much better than its accuracy, i.e., close to the camera's sensitivity. An IR camera's sensitivity (and repeatability) is typically on the order of hundredths of a degree Celsius. Consequently, a reference with known temperature and emissivity placed in the camera's field-of-view when taking measurements can provide two potential benefits: First, measurement error can be reduced by using a calibration procedure. Second, IR imagery taken by different IR cameras during different test episodes can be directly compared.