Radiometrically Accurate Thermal Vehicle Targets for Synthetic Vi

Carlson Center for Imaging Sciencein partial fulfillment of the requirementsfor the Master of Science Degree at the Rochester Institute of Technology Abstract Collecting large scientific

in the Chester F Carlson Center for Imaging Science

of the College of ScienceRochester Institute of Technology

COLLEGE OF SCIENCEROCHESTER, NEW YORK

CERTIFICATE OF APPROVAL

M.S DEGREE THESIS

The M.S Degree Thesis of David B Rhodes

has been examined and approved by the

thesis committee as satisfactory for the

thesis required for theM.S degree in Imaging Science

Dr Zoran Ninkov, Thesis Advisor

Dr J Daniel Newman

Dr David Messinger

Date

ii

byDavid B RhodesSubmitted to theChester F Carlson Center for Imaging Science

in partial fulfillment of the requirementsfor the Master of Science Degree

at the Rochester Institute of Technology

Abstract

Collecting large scientific quality thermal infrared image and video data sets is an sive time consuming endeavor Thermal infrared imagers cost much more than comparablevisible systems and require skilled experienced operators Also, time and experienced per-sonnel are required to collect quality ground truth Often it is advantageous to performcomputer simulations as an alternative to collecting image and video data with real camerasystems As long as enough physics is incorporated into the models to give accurately com-parable results to real imagery, simulated data can be used interchangeably Generatingsynthetic images and video has the added benefit of being flexible as the user has controlover every aspect of the simulation Simulations are not subject to restrictions such aslocation, weather conditions, time of day, or time of year Ground truth is assigned instead

expen-of measured in the synthetic world so it is known a priori This thesis illustrates a method

of using the Digital Image and Remote Sensing Image Generation (DIRSIG) software tocreate simulated infrared images and video of validated thermal target vehicle models in-side thermal infrared wide-area scenes A finite difference heat propagation and surfacetemperature solver, ThermoAnalytics Multi-Service Electro-optic Signature (MuSESTM),

iii

was used to accurately model the emissive thermal target vehicles Validation of the mal target vehicle model was performed using images taken from a laboratory calibratedMWIR camera Images taken with the calibrated camera of the same type of vehicle asthe target model were compared to the synthetic images for the same conditions for vali-dation Target vehicle motion was added to the simulations through the use of Simulation

ther-of Urban Mobility (SUMO), DIRSIGs movement files, and custom python scripting Theoutput images from DIRSIG were then laced together into video The resulting video wasused to test three tracking algorithms illuminating each one’s strengths and weaknesses

I would like to thank my thesis committee members: Dr Zoran Ninkov, Dr Dan Newman,and Dr Dave Messinger My research has been guided by your support and guidance.Thank you Zoran for getting me involved with DIRSIG while I was still at U of R Thanks

to CEIS and ITT (now Harris Corporation) for funding my research Without funding Iwould not have been able to finish Thank you Dan Newman, Paul Lee, and Greg Gosianfor offering suggestions on how my research should proceed from a programmatic point ofview Thank you Ross Robinson, Kenny Fourspring, and Kyle Ausfeld for being in the lab

to bounce questions off and helping with my numerous data collections Thanks you JudyPipher and Craig McMurtry from U of R for helping me to understand infrared detectorarrays and offering your camera system and truck for data collections Thank you RolandoRaque˜no, Andy Scott, Scott Brown, Adam Goodenough, Niek Sanders, Mike Gartley, andDave Pogorzala for answering my many DIRSIG questions Thank you Mike Presnar forhelping me get a good start on making video in DIRSIG Thank you to Jason Faulringfor helping me with the digital interface to the KIR-310 camera and operating WASP andWASP Lite on data collections Thank you Nina Raque˜no for loaning me temperaturemeasurement equipment and measuring materials for me Thank you Dr Schott forteaching me about remote sensing and for digging into the vault for old THERM validationdocumentation that I’m sure do not exist anywhere else Thank you Brett Matzke, JaredHerweg, and Kenny Fourspring for keeping me sane The many bike rides and times wehung out kept me from imploding under the stress of being a single parent and a full-timegraduate student Thank you Diane Alsup for proof reading this thesis and spending yourlife with me You are all very much appreciated

v

you have grown to become.

vi

The author was born on September 23, 1974 in Elmira, NY to Gerald and Regina Rhodes.

He attended Corning Community College part-time from 2000 to 2004, and graduatedhighest program GPA with an Associates in Science He attended SUNY Brockport from

2004 to 2006 graduating Cum Laude with a Bachelor of Science dual degree in Mathematicsand Physics He attended University of Rochester from 2006 to 2009 completing thePhysics Coursework before leaving to attend RIT from 2009 to 2012 The author from

2012 to the time of the publishing of this thesis is employed as a Senior Image Scientistworking for Harris Corporation, Rochester NY

vii

Acknowledgements v

2.2 Radiometry 10

2.2.1 Introduction 10

2.2.2 Radiometric Quantities 11

2.2.3 Blackbody Radiation and Emissivity 13

2.3 Atmospheric Transmission 14

2.4 Heat Transfer 16

2.5 Remote Sensing 17

3 Software 20 3.1 DIRSIG 20

3.1.1 Introduction 20

3.1.2 Scene Geometry 23

3.1.3 Materials 25

3.1.4 Emissivity 26

3.1.5 Extinction 27

3.1.6 Atmosphere 28

3.1.7 Platform 28

3.1.8 Platform Motion 29

3.1.9 Tasks 30

3.1.10 DIRSIG’s Radiometry Solvers 30

3.1.11 DIRSIG’s Thermal Solver (THERM) 30

3.1.12 External Thermal Model (ThermoAnalytics MuSESTM) 31

3.2 MODTRAN 31

3.3 MuSESTM 32

3.4 SUMO 33

4 Procedure 36

4.1 Scene 36

4.2 Target Model 38

4.3 Motion 42

4.4 Video 44

5 Truth Data 46 5.1 Camera Calibration 46

5.2 Data Collection 50

6 Results 52 6.1 Scene Comparison 52

6.2 Target Validation 54

6.3 Videos 55

6.4 Target Tracking 58

7 Summary and Future Work 60 7.1 Summary 60

7.2 Future Work 61

1.1 Frame of SENSIAC ATR LWIR video of vehicle moving across field of view 4

1.2 Frame of DARPA VIVID video of vehicle moving through intersection 5

1.3 Sample image rendering of warehouse scene by DIRSIG 6

1.4 Sample image rendering of military scene by CameoSIM 9

2.1 Plot of Radiance vs Wavelength of a 293K (20◦C) Blackbody Note peak

3.2 Surface temperature plot of target vehicle as modeled by MuSESTM 33

3.3 Screen capture of SUMO GUI showing cars driving on road network 35

4.1 DIRSIG Rendering of Megascene #1 Depicting Suburban Rochester, NY 37

4.2 Plot of three emissivity curves for asphalt from ASTER Spectral Library 38

4.3 MuSESTM surface temperature plot showing fully modeled exhaust 39

xi

4.4 Image from KIR-310 camera of target vehicle hood temperature ments 40

measure-4.5 Visualization showing original and repaired movement file Axis scales arepixels relative to image origin, where pixel size is 0.15 meters square 44

5.1 Flat field image from KIR-310 camera calibration Note that array is readout from four output channels resluting in four vertical bands of differentdigital count Axis scales are on pixels (640 x 486) 47

5.2 Kodak Research Laboratory’s KIR-310 MWIR camera 48

5.3 Plot of digital count vs radiance for KIR-310 camera Red dots are surements and blue line is line fit Note near perfect linearity 49

mea-5.4 Image from KIR-310 camera showing the target vehicle (2004 Nissan Titan)and two calibration panels; light panel is black felt for lens self emissioncalibration, and dark panel is copper plate for accurate ambient temperaturemeasurement Note bright areas near wheel wells and grill as well as tworegions of the hood 51

6.1 Comparison between WASP image from SHARE 2010 data collection (left)and DIRSIG rendered background image (right) Both images are of sub-urban Rochester, NY under similar collection conditions 53

6.2 Comparison Between ITT Exelis WAPS Image (left) of urban Rochester,

NY and DIRSIG rendered background image (right) Both images representsimilar collection conditions 54

6.3 Radiometrically calibrated KIR-310 image of thermal target vehicle (2004Nissan Titan) Axis scales are pixels (640 x 486) Gray bar shows radiance

in W/m2· sr 56

6.4 Comparison Images; calibrated truth image (left), difference in absolutevalue (center), and synthetic target model rendered in DIRSIG (right).Gray bars show radiance in W/m2 · sr Note that difference gray bar is

an order of magnitude less than truth and synthetic image gray bars 57

6.5 Frame of LWIR synthetic video with target vehicle driving on streets asseen outlined in red at top center 57

6.6 Time series of results of MS, AKF, and PFAKF tested on synthetic video.Increasing time from left to right with target vehicle traveling down Wheneach algorithm is relying on measurement the box is red and when it relies

on prediction it turns yellow 58

6.7 Time series of results of MS, AKF, and PFAKF tested on synthetic video.Increasing time from left to right with target vehicle traveling down Wheneach algorithm is relying on measurement the box is red and when it relies

on prediction it turns yellow 59

A.1 Plot of path in mov file for vehicle 56 overlayed on Megascene #1 textureimage Note that red dot denotes vehicle starting position Axes are pixels

in units of image ground sampling distance of 0.15 meters 64

A.2 Zoomed in Plot of Corner for Vehicle 56 Cutting of corner with no smoothradial transition is evident Axes scales are in pixel units of GSD of 0.15meters 65

A.3 Diagram of geometry for corner repair procedure Blue lines are the originalmotion track while dashed line depicts the corrected trajectory 67

A.4 Visualization showing original and repaired movement file White dots ignate the new 0.1 second interpolated vehicle locations 69

des-A.5 Plot of 0.1 second interpolated corner of Vehicle 56 70

2.1 Radiometric Quantities 11

2.2 Molecular weights from US Standard Atmosphere 1976 16

4.1 November 2009 Camera Systems[1] 41

xiv

u Energy Density 11

Q Radiant Energy 11

Φ Radiant Flux 11

M Radiant Exitance 11

E Irradiance 11

Ω Solid Angle 12

r Radius 12

θ1/2 Cone Half Angle 12

L Radiance 12

dA Differential Area Elememt 12

dAprojected Projected Differential Area Element 12

π Magic Pi 12

Mλ,BB Blackbody Spectral Radiant Exitance 13

Mλ Spectral Radiant Exitance 13

T Temperature in Kelvin 13

ǫ Spectral Emissivity 13

x x-Coordinate 16

y y-Coordinate 16

xv

z z-Coordinate 16

Iin Internal Generated Heat Function 16

C Specific Heat Capacity 16

t time 16

λx Thermal Conductivity in x-direction 16

λy Thermal Conductivity in y-direction 16

λz Thermal Conductivity in z-direction 16

Ld Direct Solar Radiance 18

Ldn Downwelled Radiance 18

Lup Upwelled Radiance 18

Lb Background Radiance 18

Es,λ Spectral Solar Irradiance 18

σ Solar Incidence Angle Measured From Surface Normal 18

τ1 Atmospheric Transmission from Sun to Target 18

λ Wavelength 18

rλ Spectral Reflectivity 18

LT,λ Direct Emissive Spectral Radiance 18

F Exposed Sky Function 18

Ed,s,λ Downwelled Solar Spectral Irradiance 18

Ed,ǫ,λ Downwelled Emission Spectral Irradiance 18

rd Diffuse Reflectance 18

Lb,s,λ Solar Background Spectral Radiance 18

Lb,ǫ,λ Emissive Background Spectral Radiance 18

τ2 Atmospheric Transmission from Target to Sensor 18

Lu,s,λ Upwelled Solar Spectral Radiance 18

Lu,ǫ,λ Upwelled Emissive Spectral Radiance 18

be distinguished one from another by their thermal emission Differences in emissivity ofobjects in thermal equilibrium can also be distinguished from one another This has beenused to image thin oil films floating on the surface of the ocean[2].

In the realm of defense, security, and sensing target tracking has proved to be of most importance The development of good tracking algorithms for persistent surveillance

ut-1

requires sound testing in a variety of scenarios To be truly persistent the tracking must

be able to be performed both day and night in all types of weather conditions This quires sensing at more wavelengths than the visible part of the electromagnetic spectrum.Thermal infrared lends itself to persistent surveillance because the signature of a targetchanges very little with differing illumination Since all objects above absolute zero emitlight and targets around 300K emit that light predominately in the thermal infrared, peak-ing around 10µm, they can be readily seen at night by using a thermal imager This willsimplify instruments to only using a thermal infrared camera for persistent surveillance.Relatively small amounts of calibrated thermal infrared images and video exist Evenless can be readily available to perform research on There are many reasons the thermalinfrared data is so scarce It is assumed that much of the existing data has restricted ac-cess due to International Traffic and Arms Restrictions (ITAR), proprietary concerns, andsecurity clearance Two data sets that are available are the Automatic Target Recognition(ATR) Algorithm Development Image Database from Military Sensing Information Anal-ysis Center (SENSIAC)[3,4] and Defense Advanced Research Projects Agency’s (DARPA)Video Verification of Identity (VIVID)[5, 6, 7] In addition to the scarcity of calibratedthermal video, it is often too expensive and time consuming to collect in large quantitiesfor academic research

re-Thermal infrared camera systems are currently much more expensive than comparablevisible systems There is relatively low commercial demand for scientific grade infraredimaging systems when compared to the consumer digital camera market Thermal infraredcamera systems also use exotic detector materials such as HgCdTe[8] and InSb[9], becausetheir bandgap makes them sensitive to light in that region of the electromagnetic spectrum.Visible cameras use standard silicon, which has benefited from many years of improvementand cost reduction in the semiconductor industry Bump bond hibridization of infrared

focal plane array detectors to silicon readout circuitry required for detector materials that

do not lattice match silicon decreases the yield and therefore increases the cost Coupled Devices (CCDs)[10] and Complementary Metal-Oxide-Semiconductor (CMOS)arrays, the two most common visible technologies, may benefit from hybridization, butthey do not require this process[11]

Charge-The proposed solution to the limited number of data sets is to generate comparablesynthetic video Computer generated synthetic video has many advantages over videocollected in the real world Once real world video has been collected it is difficult to adaptfor depiction of different scene conditions and scenarios For example, if it is desiredthat a target vehicle turn at a given intersection this cannot easily be changed afterthe video is collected This change is trivial in software for generating synthetic video.Good truth is also arduous to obtain in the real world In the synthetic world truth isassigned and therefore known a priori Weather and atmospheric conditions are fixedfor a given real world video collection Such conditions are inputs to the models used increating the synthetic video and can be changed to produce entirely different simulations

if necessary Proper thermal video data collection requires multiple people controlling theimagers, gathering truth data, and coordinating logistics This is all but eliminated whenproducing synthetically generated video As long as there is computing power available,

a single person can create a very large data set

1.2 Previous Work and Existing Methods

The creation of thermal infrared image and video data sets has been an endeavor metwith many difficulties Each data collect requires good ground truth and proper imagingequipment Simulation requires extensive knowledge, software, and compute power toget any meaningful results With growing interest in using thermal infrared imagers for

persistent surveillance application many have tried to develop methods of building largedata sets of thermal video upon which to test algorithms The ATR data from SENSIACwas developed for military thermal tracking research It is an extensive video data set ofmany vehicles moving at a distance from a thermal imager at a horizontal wide-area view.

A frame from one of the ATR videos can be seen in Figure1.1 The drawback to the ATRdata is the scene is wide open and there is only one look angle per video Also, it lacksthe flexibility and absolute truth gained from synthetic images and video

Figure 1.1: Frame of SENSIAC ATR LWIR video of vehicle moving across field of view.DARPA’s VIVID project includes many wavelength regions including three videos in

the thermal VIVID videos are taken from airborne platforms that image vehicles driving

on streets and roads Although this makes the videos more useful than the ATR data fortesting persistent surveillance airborne traker applications, it still has the same weaknesses

as the ATR data when it comes to flexibility and truth A frame from one of the VIVIDvideos is shown in Figure1.2

Figure 1.2: Frame of DARPA VIVID video of vehicle moving through intersection.There are also some efforts to simulate physically accurate thermal infrared imageswith software The Digital Image and Remote Sensing group at RITs Chester F CarlsonCenter for Imaging Science has been continuously developing a tool to do this for over

twenty years The Digital Image and Remote Sensing Image Generation (DIRSIG) is

a first principles synthetic image generation software that produces images of aperturereaching radiance A sample image rendering is shown in Figure1.3 DIRSIG is described

in more detail later in Section 3.1 DIRSIG has been proven useful for thermal infraredpolarimetric modeling[12]

Figure 1.3: Sample image rendering of warehouse scene by DIRSIG

DIRSIG alone was not designed to produce images with objects that have internallygenerated heat Such objects have complicated processes that require external thermalpropagation models to generate accurate surface temperature plots.

The US Air Force’s Infrared Modeling and Analysis (IRMA)[13,14] software has beendeveloped to generate imagery simulating sensors from the visible, infrared, millimeterwave radar, and Synthetic Aperture Radar (SAR) IRMA is currently being developedand maintained by Northrup Grumman Similar to DIRSIG, IRMA assigns materials tofacetized geometry The material assignment references reflectance, thermal properties,and texture files IRMA was designed for image fusion and has three channels, the pas-sive, radar, and lidar channels The passive channel uses a subprogram called ENVIROthat computes one-dimensional heat transfer of targets and backgrounds Then ENVIROproduces calculated facet temperatures for the passive image generator Due to the simplethermal model, IRMA is not well suited for creating realistic high fidelity thermal imagesand video required for this work

Lockheed Martin’s UK division developed a software package called CameoSIM thatcreates images from first principles similar to DIRSIG CameoSIM has an atmosphericdatabase which incorporates direct solar and lunar flux, sky flux, and path radiance aswell as transmission into the thermal environment[15, 14] A synthetic image is renderedincorporating the effects of the sensor and the atmosphere by observing a scene A samplerendering from CameoSim is shown in Figure 1.4 CameoSIM has been shown to haveproblems capturing glints from curved surfaces in the MWIR unless high facet countgeometry has been used High facet counts increase computation time[16] CameoSIM isnot available for use in the United States

1.3 Approach

This thesis discusses the software and data the author used to not only generate syntheticthermal video, but also outlines a flexible workflow that can be leveraged to generatevideo for any number of conditions Proper thermal target tracking video requires cor-rectly modeled background, target geometry, target signature, and target motion Forthe background geometry and thermal properties the Digital Image and Remote SensingImage Generation (DIRSIG) tool along with the thermal signature prediction and analysistool, THERM, was used The background scene modeling is described in Section4.1 Togenerate a proper thermal vehicle target’s geometry and thermal signature, the Thermo-Analytics MuSES software was used The target model was validated against collectedthermal imagery of a similar vehicle using laboratory calibrated imaging systems Thethermal target vehicle model is described in Section 4.2 The motion of the vehicles inthe scene was simulated using SUMO Sumo has limitations that needed to be overcomefor remotely sensed target tracking purposes The motion data from SUMO was adjustedand improved using geometry techniques developed for this research The motion im-provements are discussed in Appendix A The vehicle target motion modeling as well asthe confuser vehicles’ motion are described in Section 4.3 To generate the video, mul-tiple frames were generated using the above techniques in parallel on a compute clusterenvironment The resulting frames were processed and laced together into video usingFFMPEG and Python code The video generation process is described in Section4.4andthe Python code is shown in AppendixB

Figure 1.4: Sample image rendering of military scene by CameoSIM.

Background Information

2.1 Introduction

Before describing the research performed for this thesis important background informationmust be presented To simulate the real world accurately in the infrared, one must have abasic understanding of radiometry The most important radiometric quantity is radiance

A general understanding of how light propagates through the atmosphere is essential Aknowledge of heat transfer is required for understanding the physics behind the targetmodel Finally, since the simulation is to replace real-world remotely sensed data, someremote sensing terminology is also important

etry In radiometry, certain assumptions are required All calculations are performed as

if to obey geometrical optics, the sources are incoherent, and energy is conserved derstanding radiometry is very important for imaging in the MWIR and LWIR as theseregions of the electromagnetic spectrum are dominated by thermal emission.[17,18]

Un-2.2.2 Radiometric Quantities

Radiometry is based on a set of radiometric quantities that have a standard definition perthe CIE They can be found in Table2.1below The most useful quantity is radiance Forthis thesis the term aperture reaching radiance refers to the radiance which is measured

at the first surface aperture of an imaging system [19,20]

Quantity Symbol Definition Units

a surface The remaining quantities in Table2.1involve the concept of flux per unit solid

angle The solid angle Ω is defined as the area a enclosed by a cone intersecting a sphere

of radius r divided by r2 Another way to define the solid angle is by the cone half angle

θ1/2 as seen in Equation2.1

Radiance L is the flux per unit area projected per unit solid angle leaving a referencesurface Radiance can be defined by using the angle θ between the surface normal and thearea element dA by Equation2.2

E = Icos(θ)

Since flux is a spectral quantity all of these quantities are spectral in nature That is tosay, the radiance leaving a surface is wavelength dependent If the surface is consideredLambertian, a perfect diffuse reflector or emitter, then the radiance and radiant exitance

is related by the Magic π [21] as shown in Equation2.4

This concept of a Lambertian surface becomes important in Section 2.2.3 as blackbodysurfaces are also Lambertian

2.2.3 Blackbody Radiation and Emissivity

The amount of radiation emitted from the surface of an object is measured relative to

a hypothetical perfect emitter referred to as a blackbody A blackbody has the uniqueproperties that it is a perfect absorber as well as a perfect emitter Blackbody radiation

is governed by Equation2.5 referred to as the Planck equation

Mλ,BB(T ) = 2πhc

2

λ5

1

a 293K (20◦C) blackbody is shown in Figure 2.1 Note that the peak of the spectralradiance is at a wavelength of 10 µm illustrating why the long wave infrared is importantfor persistent surveillance

Since no true blackbodies exist in nature, real materials can only be compared toblackbodies The emissivity ǫ of an object can be thought of as the factor which determineshow similar to a blackbody that object behaves This relationship is seen in Equation2.6.The spectral radiance from a thermally emissive object at a given temperature can bedetermined by knowing its spectral emissivity Assuming the object is in steady-statethermal equilibrium and opaque (i.e., non-transmissive) the emissivity can be determined

by Kirchhoff’s law to be one minus the objects reflectivity

ǫ(λ) = Mλ(T )

Spectral emissivity is a unitless factor ranging from zero to one

Figure 2.1: Plot of Radiance vs Wavelength of a 293K (20◦C) Blackbody Note peak at10µm.

2.3 Atmospheric Transmission

Due to evolution, human beings have the good fortune of seeing in the visible part ofthe electromagnetic spectrum, approximately 390 to 750 nm The atmosphere is almostcompletely transparent in this region; which is not so for most of the rest of the spectrum.Molecular absorption and scattering creates transmission windows framed by regions ofdeep absorption where the atmosphere is opaque In the thermal infrared there are twomain transmission windows, 3 to 5µm known as the Mid-Wave Infrared (MWIR), and

8 to 14µm known as the Long-Wave Infrared (LWIR) A plot of simulated atmospherictransmission can be seen in Figure 2.2 Although the transmission in these two regions

is not perfect, it is high enough to produce good image results The largest contributers

to the absorption of electromagnetic radiation in the infrared as it passes through theatmosphere is water and carbon dioxide Water is also the most variable depending onlocation, season, and time of day

Figure 2.2: Plot of MODTRAN generated atmospheric transmission typical for Rochester,NY

The atmosphere is made up of many layers consisting of different quantities of ular constituants The most dominant are nitrogen, oxygen, argon, and carbon dioxide.The molecular weight and fractional volume of each of these molecules are seen in Table

molec-2.2taken from the U.S Standard Atmosphere of 1976[22]

Gas species Molecular weight Fractional volume

heat generation, Iin, is a function that describes the amount of heat generated internallyfor a given object.

Thermal convection is a complex phenomenon where heat is transferred through themeans of an intermediary fluid, usually air or water The fluid warms up while in contactwith a warm surface causing it to expand Expansion makes the fluid less dense andtherefore more boyant than nearby cooler fluid This induces a flow as more dense fluidseeks to take the place of the warmer fluid as it moves away Defining the flow equations

is difficult due to the complexities of fluid dynamics Convection models require a goodComputational Fluid Dynamics (CFD) software model The mathematical description ofconvection is beyond the scope of this thesis

Radiation is the transfer of electromagnetic energy carried by photons emitted fromall objects above absolute zero In this thesis only photons emitted from objects between

300 Kelvin and above will be considered, because these objects blackbody radiation peaks

in the wavelength regions of interest Radiation is described in Section2.2

computer simulations.

For the purpose of remote sensing there are two regions of the electromagnetic spectrumthat behave very differently, the solar reflected dominated and the thermally emissivedominated The solar reflective region ranges from across the entire visible to the MWIR.The thermally emissive region extends across the MWIR and the LWIR It is important

to notice that the MWIR is a cross-over region where both the solar reflected light andthermally emitted light are non-negligible.That is not to say that the LWIR is uneffected

by solar radiation indirectly, but direct solar reflection is not appreciable in that region

As discussed in Section 2.2, radiometry assumes that energy is transferred in straightpaths obeying geometric optics All radiation a sensor receives can be classified into eightways Radiation can come directly, downwelled, upwelled, and from the background Each

of these can either be from the sun or emitted from the scene The direct radiance, Ld, is

a combination of the reflected sunlight and the self emitted light coming from the target ofinterest to the sensor The downwelled radiance, Ldn, is the additional radiation comingfrom the target that was either scattered or emitted from the atmosphere The upwelledradiance, Lup, is the radiance reaching the sensor, which is either scattered or emitted fromthe atmosphere Upwelled radiation can lead to hazy appearing images The backgroundradiance, Lb, is the radiance coming from objects in the scene that are not the target andmay not be in the field of view A tree or building can be an example of a backgroundobject Combining all eight radiance terms together to give the total aperture reachingradiance forms Schott’s big equation[21] as seen in Equation 2.8

Lλ= [Esλcos(σ)τ1(λ)r(λ)π + ǫ(λ)LT λ+ F (Edsλ+ Edǫλ)rd (λ)

π +(1 − F )(Lbsλ+ Lbǫλ)rd(λ)]τ2(λ) + Lusλ+ Luǫλ

(2.8)

The function F defines the portion of the scene that is illuminated by the sky directly

and not the portion that is obscured by the background.

soft-20

to provide calibration The current version of DIRSIG is DIRSIG 4 DIRSIG is reversecompatible with the previous versions and is under constant development Some of theinput files used by DIRSIG 4 originated from DIRSIG 3 The DIRSIG 4 simulation inputfiles begin with the simulation (.sim) file The DIRSIG simulation file contains paths tofive input description files These files are the scene description file (.scene), atmosphericconditions description file(.atm), the platform description file (.platform), the platformposition file (.ppd), and the data collection tasks file (.tasks).

This scene description incorporates geometries of objects such as terrain, trees, ings, and vehicles with material databases, emissivity files, extinction files, texture maps,

build-as well build-as other optional material property files The objects are described in ObjectDatabase (.odb), Geometric Database (.gdb), and Geometry List (.glist) files At a baselevel the gdb files are three-dimensional (3D) Computer-Aided Design (CAD) files in hu-man readable text format The gdb is broken up into Objects, Parts, and Facets, witheach facet defined by three or four vertices and a normal vector Each facet is assigned amaterial identification corresponding to a specific material in the material database (.mat)where the material properties are assigned The odb and glist files describe an object’s ge-ometry by linking one or more gdb files together The odb and glist are human readabletext and XML respectively

The atmospheric description uses an external atmospheric radiative transfer model veloped by the United States Air Force called MODerate resolution atmospheric TRANs-mission (MODTRAN) For a more complete description of MODTRAN see section 3.2.DIRSIG does not call MODTRAN directly for each ray cast that passes through the atmo-sphere Instead the DIRSIG distribution contains the make adb program that calls MOD-TRAN a number of times to create a look-up table (LUT) called atmospheric database(.adb) To use make adb, a MODTRAN tape5 file (.tp5) must exist with the specifics

de-of the chosen atmosphere set Then make adb uses the tp5 file to run MODTRAN toproduce the adb file The adb file contains spectral irradiance and transmission datafor the Sun and Moon, upwelled and downwelled radiance for a sampling of directionsbased on the sensor model geometry as well as the radiometry solver being used Duringexecution DIRSIG reads from and interpolates data from the adb file to produce a goodrepresentation of the effects of the atmosphere on each ray cast Weather history for up to

48 hours is needed in order for DIRSIG’s thermal model, THERM, to predict the ature of each point in the scene for any time during a 24 hour day The weather historyfile (.wth) is an input file with tabular ascii data that contains 13 columns describing theweather at one quarter hour time increments The wth file includes data such as the airtemperature, atmospheric pressure, relative humidity, dew point, and precipitation.The platform description file (.platform) is a description of the simulated camera The.platform file contains the information about the simulated detector’s spectral bandpasses,responses, array shape and array size The output image is also defined in the platformdescription By default DIRSIG outputs an Harris Visual Information Solutions Environ-ment for the Visualization of Images (ENVI) image data cube

temper-The platform motion description is contained in the Platform Position Database (.ppd)and can either be static or dynamic A static platform motion places the camera modelfrom the platform file at a location in the 3D space pointing it in a specific direction Adynamic platform motion places the camera model at a location and pointing, then movesthe location and pointing along a predefined path The dynamic platform simulates amoving platform such as an airborne or spaceborne platform

The data collection tasks (.tasks) file is simply a file that defines the type of imagecapture, whether a single instantaneous capture or a continuous capture over a window oftime The date and time are also specified in the tasks file