Weather Effects on Target Acquisition

8/13/2019 Weather Effects on Target Acquisition

http://slidepdf.com/reader/full/weather-effects-on-target-acquisition 1/55

Army Research Laboratory

Weather Effects on Target Acquisition

Part I : Sensor Performance Model

Infrared Algorithms

by

Richard C. Shirkey

Barbara J. Sauter

Computational and Information Sciences Directorate

Battlefield Environment Division

Rene V . Cormier

U.S. Air Force Research Laboratory

(Dynamics Research Corporation)

ARL-TR-821 uly 2001

Approved for public release; distribution unlimited.

2 0 0 1 0 8 0 2 3 4



NOTICES

Disclaimers

The indings n his eport re ot o e onstrued s n fficial Department f he Army osition, nless o esignated y ther

authorized documents.

Citation of manufacturers' or trade names does not constitute an official

endorsement or approval of the use thereof.



REPORT DOCUMENTATION PAGE Form Approved OMB No 0704 0188

Public reporting burden for this collection of information is estimated to average our per response, including the time for reviewing instructions, searching existing data sources,

gathering andmaintaining the data needed, and completing and viewing the collection information. end comments regarding this burden estimate or any other aspect of this collection of

information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway,

Suite 1204, Arlington, V A 22202-4302, and to the Office of Management and Budget, Paperwork Reduction (0704-0188), Washington, DC_20503_

1. AGENCY US E ONLY Leave blank) 2. EPORT DATE July 2001

3. EPORT TYPE AN D DA TES COVERED FINAL

4. TITLE AN D SUBT ITLE Weather ffects n arget Acquisition art : Sensor erformance Model nfrared

Algorithms

6. AUTHOR(S) Richard C. Shirkey

Barbara J. Sauter Rene V. Cormier, U.S. A ir Force Research Laboratory, Dynamics Research Corporation

7. PERFORMING ORGANIZATION NAME(S) AN D ADDRESS(ES) U.S. Army Research Laboratory Information Science and Technology Directorate

Battlefield Environment Division

ATTN: AMSRL-CI-EW White Sands Missile Range, NM 88002-5501

9. SPONSORING/MONITORING AGENCY NAME(S) AN D A DDR ESS(ES )

5.

UNDING

N UMBER S

8. ERFORMING ORGANIZATION REPORT N UMBER

ARL-TR-821

U.S. Army Research Laboratory

2800 Powder Mill Road

Adelphi, M D 20783-1145

1 0. SP O N SO R IN G / MONITORING A G EN C Y REPORT N UMBER

ARL-TR-821

1 1 . SUPPLEMENTARY NOTES

12a . ISTRIBUTION/AVAILABILITY STA TEMEN T Approved fo r public release; distribution unlimited.

12b . ISTRIBUTION C O DE A

13 . ABSTRACT Max imum 20 0 words) The U.S. A ir orce, Navy, nd Army re n he rocess f upgrading he lectro-Optical actical Decision A id (EOTDA). he EOTDA has been used to predict the impact of weather and time of day on target acquisition. he

upgraded rogram is alled he Target Acquisition Weather Software TAWS). ew eatures f the A W S will

include: automated data access; upgraded path radiance routines; replacement of separate infrared IR); television (TV) and night vision (NV) sensor performance models with Acquire; and a U.S. Army standard sensor performance

model, which has become a standard in Department of Defense fo r IR , TV, and NV systems. o quantify the effects

on redicted arget cquisition ange f upgrading he ensor erformance model, omparison f he AW S

Version 2 with Acquire has been undertaken. Weather effects on target acquisition are examined also insome detail.

1 4. SUBJEC T TER MS weather effects, target acquisition, Acquire, TAWS, clutter

1 7. SEC UR ITY CLASSIF ICATION OF REPORT

UNCLASSIFIED

18 . SECURITY CLASSIF ICATION OF THIS PAGE UNCLASSIFIED

1 9. SECURITY CLASSIF ICATION OF A BSTR A C T UNCLASSIFIED

1 5. UMBER OF PAGES 64

1 6. RICE C O DE 2 0. IMITATION O F A BSTR A C T

SAR

NSN 7540-01-280-5500 Standard Form 29 8 (Rev. 2-89) Prescr ibed by ANSI Std.Z39-18



Acknowledgements

The uthors would ike o cknowledge he many nd eneficial discussions with Dave Dixon, Training and Doctrine Command Analysis

Center-White ands Missile ange TRAC-WSMR), John Mazz, Army

Materiel Systems Analysis Agency (AMSAA), Melanie Gouveia, Logicon,

and av e chmieder, eorgia ech esearch nstitute GTRI) hat

occurred uring he ourse f his ork. heir id n roviding information and tracking down documents concerning both the Acquire

and TAWS sensor performance models was invaluable.



Contents

Preface iii

cknowledgements V

Executive Summary 1

1. ntroduction 5

2. ackground 7

2.1 Minimum Resolvable and Detectable Temperatures 7

2.2 ensor Performance Models 8

2.2.1 Acquire SPM and Methodology 9

2.2.2 AW S SPM and Methodology 14

2.2.2.1 Resolved Versus Unresolved Targets 16

2.2.2.2 Determination of SC R 17

2.2.3 Discussion 21

3 . Comparisons 23

3.1 cenarios 23

3.2 Model Runs 24

3.2.1 haracteristic Dimension 24

3.2.2 ensor Curves 26

3.2.3 Aspect Ratios 26

3.2.4 Backgrounds 27

3.2.5 Clutter 28

3.3 esults 29

3.3.1 Clutter/Complexity Effects 30 3.3.2 Weather Effects 33

3.3.2.1 Visibility 33

3.3.2.2 elative Humidity 33

3.3.2.3 ky Cover 36

3.3.2.4 og and Precipitation 37

vu



Figures

Tables

4. onclusions 9

References 1

Acronyms 5

Distribution 7

1 . quivalent bar pattern 0

2. arget-acquisition methodology 1

3a. nitial clutter temperature algorithm 0

3b . Current clutter temperature algorithm 0 4. robability of detection versus resolution for 50 percent

acquisition level using the T A W S and Acquire algorithms 1 5a. -80B 5

5b. -80U 5

6 . Comparison of the T A W S and Acquire M R T curves 6 7. Acquire versus the T A W S clutter comparison 8 8. he TAWS versusAcquire-detection ranges at 50 percent

probability of detection 9

9. cene complexity effects on target-detection ranges 1 10 . isibility impacts onthe TAWS-detection ranges 5 11. elative humidity impacts onthe TAWS-detection ranges 5 12. ky cover impacts onthe TAWS-detection ranges 7 13. og impacts on the TAWS-detection ranges 8 14. recipitation impacts on the TAWS-detection ranges 8

1 . 50 as a function of Acquire version 4

2. ummary of detection methods 7

3. Clutter temperature C T values 8

4. Algorithm fo r determining Sc 9

5. nitial T A W S implementation of CT algorithm 9 vm



6 . Current TAWS implementation of CT algorithm 0 7. Weather conditions used in th e study 3

8. roblems between TAWS and Acquire algorithms and

their resolution 4

9. -80 dimensions 5 10. mpact of a second background on detection range 2 11. Detection ranges as a function of various weather 4

IX



Executive Summary

Overview

The range a target can be detected on the battlefield is a valuable piece of information or he attlefield ommander. etection nd ecognition

ranges epend pon he arget nd ackground haracteristics,

atmospheric ropagation, nd ensor erformance. eather actical

decision aids provide information on sensor performance under adverse

weather conditions.

The Target Acquisition Weather Software (TAWS) is an updated version

of the U.S. A ir Force Electro-Optical Tactical Decision Aid. he TAWS

provides .S. ir orce, Navy nd Army mission lanners nd

warfighters with ppropriate nformation or ptimal ensor nd/or weapon systems selection, acquisition range determination, and mission routing under degraded weather conditions. he T A W S was originally

constructed to predict detection and lock-on ranges only. ecause the U.S. Army extensively makes use of recognition and identification ranges,

methodologies fo r adding this information to the T A W S were examined. The leading algorithm, and thus, a contender fo r the replacement of the

current algorithm, is Acquire.

Physics-based actical Decision Aids TDA)s, uch s AWS, mploy

physics alculations hat have heir asis n heory r measurements.

Thus, a physics-based TDA employs routines and physics that allow it to ascertain he robability f etecting iven arget t iven ange

under xisting r redicted weather onditions. he ffects nd methodology fo r determination of th e range a target is detected under

adverse weather conditions at infrared (IR) wavelengths is the subject of

this eport. he ensor erformance model Acquire nd he enor

performance model in TAWS both use what is known as the equivalent bar attern pproach; owever, he nderlying ssumptions f ac h

algorithm re onsiderably ifferent.

he urpose f this eport is o show th e ifferences between these two methodologies and to examine

weather effects on target-acquisition ranges.



Background

Typical performance prediction models for resolved targets a target is considered o be esolved when the arget-angular ubtense nominally

exceeds he ensor's ngular ubtense n both ertical nd orizontal dimensions at the range of interest) treat the target with the bar/target

equivalency riteria nd he ensor with he minimum esolvable

temperature MRT) unction. ar - nd arget-signal quivalency s

established by quating he bar attern emperature ifference o he

target verage emperature ifference. he etection range s harply

bounded in that it can never exceed the range at which the target ceases

to be resolved, that is the (detection) range «target size/resolution.

Models fo r predicting the

detection range

of unresolved targets typically rely trictly n arget-signal trength or etection. hese models

typically abandon both the bar- and target-equivalency criteria and the M RT approach. Unresolved target models are often called hot spot or

star detection models because they rely on high-apparent contrast fo r

detection. he target is a square or circle with dimensions matched to the high temperature target area of interest. his target spot detection

methodology pplies o ases he arget s iewed gainst niform background, and detection occurs when the ignal-to-noise ratio on the

display element that subtends the target exceeds that of the background. The methodology fo r spot detection applies only to the detection of the

target (its discrimination from the background), and not to levels of target discrimination. he ensor unction s he inimum etectable temperature (MDT).

The D T s nly ppropriate or argets gainst niform r

unstructured background, fo r example, aircraft against a clear or overcast sk y r vehicles n a esert background. earching fo r anks gainst a varied terrain background requires the M R T approach. Additionally, the

M D T pproach nly epresents etection whereas he M RT pproach,

which may also be used fo r detection, is required fo r target recognition

and identification.

Target detection, recognition, and identification methodology applies to

situations in which the target is embedded in a non-uniform or cluttered background and it is necessary to separate the target characteristics from

the background. he target discrimination M R T methodology, based on

the Johnson cycle criteria in Acquire and Schmieder's criteria in TAWS, ca n be used fo r the prediction of target-acquisition range at



Conclusions

discrimination evels f detection, recognition, and dentification. oth

the TAWS and Acquire can use M DT to predict detection range also.

T he AW S urrently ses chmieder's work mplemented o redict

detection; igher isaimination evels, uc h s ecognition nd

identification, re not included. ecause Acquire s urrently both the U.S. Army and A ir Force standard fo r target acquisition, and it predicts

ranges or iscrimination evels f etection, ecognition, nd

identification it was decided to replace the urrent sensor performance

model SPM) esident n he AW S with cquire. owever, his

replacement aises he uestion f what ifferences may rise ue o

different methodologies between the SPMs in the T A W S andAcquire. o

answer his uestion or R ensors, omparison f tatic arget

discrimination methodologies and th e resultant target-acquisition ranges

produced by the TAWS and Acquire was undertaken. t should be noted

that this comparison in no way should be construed as a validation of the target-acquisition anges. ather we re xamining what, f ny,

differences that arise due to the underlying SPMs and the methods that

they implemented.

To ompare hese wo omplex arget-acquisition models equired

standardization of as many parameters as possible. o accomplish this,

one weather scenario was usedin conjunction with one sensor and target,

both with ixed rientations. hus, winter cenario was hosen nd examined using n xercised -80 oviet main battle ank against two

backgrounds vegetation and snow) at IR wavelengths. he sensor and tank were aligned such that the sensor always had a frontal view of the

tank; the ensor height was fixed at 300 t facing north. he date was

fixed t 21 December t a ocal time f 12N; the ocation was ixed t

latitude f 37°32' N, ongitude f 127°00' Seoul, South Korea). he weather onditions ncluded lear kies with arying isibility nd

relative humidity and overcast skies with varying visibility and relative

humidity. dditional ases ere un ncluding ight/heavy og

conditions, snow, drizzle, and rain.

The PM in the TAWS is based on Schmieder's image-based work and

thus requires additional algorithms to determine the effects of clutter in

the scene; these algorithms ar e not necessarily intuitive. ince Acquire's

target ransform robability unction nd chmieder's etection

probability s unction f esolution, gree easonably well nder



moderate lutter onditions, nd because Acquire s n ndustry, U.S.

Army and A ir Force standard, it was decided to replace the current SPM

with Acquire. Clutter levels can be accommodated in Acquire by varying the N50 parameter.

The incorporation of the Acquire SP M into the T A W S is scheduled fo r the

TAWS Version 3, which will be released fo r use in 2001. he comparisons

shown here, of target-acquisition range output from the current version

of T A W S with output rom Acquire, provide ositive eedback on the

benefits f his nhancement o AWS , while maintaining xisting

interfaces nd providing comparable arget-detection range predictions.

T he rimary benefit f this nhancement will e he bility o pecify

target acquisition discrimination levels, including detection, recognition,

and identification.

Because th e cases examined in this report are limited to a winter scenario in Korea, specific quantitative results of th e selected weather parameters'

impacts on target-detection range cannot be generalized to al l situations. However, these cases do highlight the importance of considering accurate atmospheric conditions in target-acquisition predictions. esults show a

smaller weather impact to Acquire detection ranges than predicted using

the current T A W S SP M under conditions of fo g or precipitation.





2.

Background

2 .1 Minimum Resolvable and Detectable Temperatures Typical performance prediction models or resolved targets a target is

considered o e esolved when the arget-angular ubtense nominally

exceeds he ensor's ngular ubtense n both ertical nd orizontal

dimensions t he ange f interest) reat the arget with he bar- and

target-equivalency criteria and the sensor with the minimum resolvable temperature MRT) unction. his methodology assumes that resolved

targets re etected, based n bserver attern ecognition. ignal strength only needs to be sufficient in order to define the pattern. he

signal s ypically efined s he emperature ifference between he

average temperature of the target and a uniform background temperature

as seen by the sensor. ar- and target-signal equivalency is established by equating he bar pattern emperature ifference o he arget verage

temperature difference. he detection range is sharply bounded so that it

can never exceed the range at which the target ceases to be resolved, that

is the detection) ange « target size /resolution. his is the maximum range t which eriodic arget an be aithfully eproduced; hus

target is considered unresolved if the projected sensor instantaneous field

of

iew

IFOV)

s

reater

han

0

ercent

f

he

arget's

ritical dimension. he percentage is taken to be 80 percent because other sensor apertures, n ddition o he detector IFOV, cause he ffective ystem

IFOV to be slightly larger.

Models fo r predicting the detection range of unresolved targets typically

rely trictly n arget-signal trength or etection. hese models typically abandon both the bar- and target-equivalency criteria and the

M RT approach. Unresolved target models ar e often called hot spot or star detection odels ecause hey ely n igh pparent ontrast or detection. he target is a square or circle with dimensions matched to the

high emperature arget rea f nterest. his arget pot etection

methodology pplies o ases n which he arget s iewed gainst a uniform background, and detection occurs when the signal-to-noise ratio

on he isplay lement hat ubtends he arget xceeds hat f he

background. hat s, ufficient mount f arget nergy eaches

detector lement o reate ot spot on he ystem isplay. For this

reason, pot etection s lso eferred o as tar etection.) The



methodology fo r spot detection applies only to the detection of the target

(its iscrimination rom he background), nd ot o evels f arget

discrimination. he ensor unction s he inimum Detectable

Temperature MDT). etection ange redictions re ot harply bounded with his nresolved arget methodology ince arget-signal

strength does not abruptly disappear after becoming unresolved.

T he DT s nly ppropriate or argets gainst niform r

unstructured background; fo r example, aircraft against a clear or overcast sk y r ehicles n a desert background. earching or anks gainst a

varied terrain background requires the M RT approach. Additionally, the

M D T pproach nly epresents etection whereas he M R T pproach,

which may also be used fo r detection, is required fo r target recognition

and identification. f a target is hot enough, the M D T approach predicts

target detection even though the target may be smaller than a forward- looking infrared (FUR) detector element. or al l practical purposes, the M DT pproach s ot sed n he Army's ombat imulations ince

recognition r dentification s equired before iring n arget. t should also be noted that in a cluttered environment the target would not

be the only hot spot.

2 .2 ensor Performance Models Target detection, recognition, and identification methodology applies to situations in which the target is embedded in a non-uniform or cluttered background and it is necessary to separate the target characteristics from

the background. he target discrimination M RT methodology, based on th e Johnson cycle criteria [4] in Acquire and Schmieder's criteria [5,6] in the TAWS, ca n be used fo r the prediction of target-acquisition range at

discrimination levels of detection, recognition, and identification. oth

the T A WS and Acquire can use M DT to predictdetection range also.

Acquire, eveloped y he .S. Army Night ision nd lectronic Sensors Directorate (NVESD), is an analytical model that predicts target- detection and discrimination-range performance fo r systems that image

in the visible, near-IR, and IR-spectral bands. anges and probabilities predicted y he model epresent he xpected erformance f n ensemble of trained military observers with respect to an average target

having a specified signature and size.



The U.S. A ir Force Research Laboratory's (AFRL) Electro-Optical Tactical

Decision Aid EOTDA) 7] was eveloped o rovide he ser with a

single iece f oftware o valuate he ombined ffects f target-to-background ontrast, tmospheric ransmission, nd ensor performance on the range at which a target ca n be detected by an imaging

device. he model treats detection by television, image intensifiers, and

thermal maging evices. he AWS, ri-Service rogram, s n

upgrade of the EOTDA.

The A W S urrently ses chmieder's work mplemented o redict

detection; igher iscrimination evels, uch s ecognition, nd

identification, ar e ot included. ecause Acquire s urrently both the

Army standard 8] nd a U.S. A ir Force standard fo r target acquisition,

and because t redicts anges or iscrimination evels f etection,

recognition nd dentification, t was ecided o eplace he urrent sensor erformance model SPM) esident n he A W S with Acquire. However, this replacement raises the uestion of what differences may

arise due to different methodologies between the SPMs in the TAWS and

Acquire. o answer this question fo r IR sensors, a comparison of static

target-discrimination methodologies and the resultant target-acquisition

ranges produced by the TAWS and Acquire was undertaken. t should

be oted hat his omparison n o way hould be onstrued s validation f he arget-acquisition anges er e. ather we re

examining what, if any, differences that arise due to the underlying SPMs

and the methods that were implemented.

2.2.1 Acquire SPM and Methodology

During he 950s, he ilitary eveloped lectro-optical mage

intensifiers, which rovided nhanced isual urveillance apabilities

under conditions of limited visibility. he complexity of these intensifiers and ssociated arget-acquisition ystems equired methodology or

evaluating performance haracteristics. ohn Johnson, of the then U.S. Army ngineer esearch nd evelopment aboratories ERDL)

(currently VESD), resented esults f xperiments with uman

observers conducted at ERDL to determine the resolution required of a system to perform certain target-interpretation processes such as target

detection and recognition. e referred to these as decision responses and said hat the processes were dependent upon the haracteristics of the optical message, he roperties f he ntensifier evice, nd he physiological response of th e human readout processes. hrough a series

of xperiments sing rained bservers ooking t argets nd ar

resolution diagrams, Johnson [4 ] developed a method relating the decision



Figure 1. quivalent bar pattern.

response o he umber f bars line airs) ormalized o he hortest

target imension hat n bserver eeded o ee o make ecision

(detection, recognition, etc.).

T he methodology developed by Johnson was simple and straightforward.

In the laboratory, scale models of various military targets were moved to

a istance where hey could just be detected as viewed by an observer

through an image intensifier. ar charts, with the same contrast as the

scale models, were then placed in the observer's field of view at the same

range s he arget. he patial requency line airs) esolved by the

observer was hen etermined s unction f ontrast. his ame

methodology was used or determining he ine airs equired by the

observers to recognize the object seen as a tank. he spatial frequency of the pattern was specified in terms of the number of lines in the pattern

subtended by the object's minimum dimension as illustrated in figure 1 .

Figure 1 shows three bars across the tank's shortest dimension as seen by an observer; this is the riterion Johnson used or ecognizing that the object was a tank. urther, Johnson found that the normalized line-pair resolution equired or articular decision esponse was early

constant fo r the group of nine military targets he employed. n the case

of target detection, that was 1.0-line pairs per shortest target dimension. These ormalized to he hortest arget imension) ine air alues

required or ecision were ound o be ndependent f contrast nd scene signal to noise ratio as long as the contrast on the bar chartwas the

same as the target contrast.

he results showed that decision levels fo r military argets may e onsidered quivalent o ar atterns f

appropriate spatial frequencies.

10



This methodology, which provided arget-discrimination riteria based

upon resolution, gained widespread acceptance within the industry and

became he ccepted riteria or erformance measurement f ptical systems. hese riteria ere eferred o s he ohnson Criteria r

equivalent bar pattern approach.

In 1974,16 years after publishing his original paper, Johnson modified his

original work nd xtended t o over R ystems 9]. his aper

emphasized hat he alues or he arious ecision evels re

representative values, essentially average values required fo r 50 percent

probability, and must not be onstrued as rigid or optimum values fo r

specific targets and target-aspect views. he values associated with the various ecision evels emained he ame, e.g., .0 ine airs or

detection) except fo r recognition, which changed slightly. his paper also

recommended and provided procedures fo r usingthe concept of MR T fo r thermal sensors. ohnson's methodology is schematically represented in

figure 2. [10]

Figure . Target- acquisition methodology.

Sensor Resolution

m i m HI nf^

mi m utfini. e U jj/lii: r

Spatial Frequency

cyrles/mrad)

Probability

ID

O jS

DMClfDI

Ibci0iriifl

£ 02.

lhrillEidc£

D IW L

The M RT s efined s he emperature ifference between uniform background nd he ars f our-bar attern, ac h bar aving

7:1 aspect ratio (so the overall pattern will be a square), which is required

by rained bserver o ust esolve ll our bars when iewing he

pattern hrough n mager. 11] he ef t side f the igure hows he

resolvable emperature ifference contrast) ersus he aximum

resolvable bar pattern (spatial frequency) as a function of contrast. or a specific target-contrast level, the maximum resolvable spatial frequency is th e ighest patial requency t which uman bserver an till

recognize he our istinct bars nd not ne r wo blobs. hus, he

temperature difference required to resolve the four bars increase, as the bars et maller. This maximum esolvable patial requency s

11



function of contrast visual or thermal) nd s the minimum resolvable

contrast (MRC) in the visual or M R T difference curve in the thermal. n

the figure, the bars represent the generic formulation, whereas the solid line would epresent he arget ontrast nd ensor esolution or

specific sensor. ith knowledge of the target's contrast (AT in the E R ) , critical dimension, range, and the atmospheric attenuation, the number of resolvable ycles, N, cross he arget's ritical imension an e

determined by

N =

f ^f->. where is the maximum resolvable patial frequency of the sensor in

cy/mr) t he pparent Ta usually n K), Hta r g s he arget-critical dimension (in meters) and R is the range (in kilometers). he apparent thermal contrast is determined by

AT a = AT T(R), 2)

where T(R) is th e atmospheric transmission. Under the assumption of a

homogeneous tmospheric path he ransmission may be ound using Beer's law

T(R) = e-ßR 3)

where s he tmospheric xtinction oefficient determined rom n atmospheric propagation code [12,13]. W e now need a way to correlate N

with the iscrimination level: etection, ecognition, and identification.

Johnson did this by establishing the so-called target transform probability

function TTPF). 9] he TPF, hown n he ight-hand ection f

figure 2, was erived rom aboratory sychophysical xperiments n which the ability of observers to perform a particular discrimination task

as a function of resolvable cycles across the target-minimum dimension

was measured. or a given discrimination task, the TTPF represents the

50-percent point, referred to as N50, as determined from this ensemble of

12



range. hus, if the range is unknown, but information is available about

the target-critical dimension, the target/background AT , and the sensor

response urve, then a olution ca n be ound through iteration fo r the range at a predetermined Pd.

A s LIRs dvanced rom irst o econd eneration, heir esolution

increased o hat here was early qual esolution long he ca n

direction and perpendicular to it. his, in part, necessitated a change in

the riginal ne-dimensional ID ) ersion f Acquire. n pdated version of Acquire, discussed in detail in reference 11 , was issued in June

1990. he modelupdate consisted of two parts:

1 . LIR90, which predicts laboratory measures and

2. two dimensional (2D) version of Acquire, which predicts field performance.

In LIR90, redicted r measured orizontal nd ertical MR T s re

averaged at a particular temperature using

/ff=(x* ) * -4

) *

This effective spatial frequency applies to the effective M RT used along

with modified alues f N50 or he ifferent iscrimination asks o predict range performance. he N50 values fo r ID or 2D are presented in table 1 . he N50 fo r a particular task, using the more recent 2D version of the model, is found by multiplying the original ID N50 values by 0.75. The amount of shift was determined by requiring the range predictions

fo r he D model o orrectly redict he esults f field ests, which

served s alidation or he riginal ID ) model. he iscrimination levels in figure 2 are fo r second generation FLIRs. s an example, if we

have a target with a critical dimension of 3 m at a range of 3 km, with an

apparent contrast that would ive a maximum resolvable frequency of 3 cycles/mrad, his would ead o robability f ecognition, or

second generation

FLIR,

of 50

percent; with a

first generation FLIR,

this would lead to a probability of recognition of 25 percent.

* The discussion in this paragraph relies heavily on reference 11 .

13



apparent contrast that would ive maximum resolvable frequency of 3 cycles/mrad, his would ead o robability f ecognition, or

second generation FLIR, of 50 percent; with a first generation FLIR, this would lead to a probability of recognition of 25 percent.

Table 1 . N5o as a function of the Acquire version Acquire version ID 2D

Detection 1.0 .75

Recognition 4.0 3.0

Identification 8.0 6 .0

2.2.2 AWS SPM and Methodology

The TAWS SPM is derived from work done in the 1980s at the U.S. Air

Force vionics aboratory now art f FRL), which ed o he development of a research grade TD A. he main difference between

the AW S nd he Avionics esearch rade ode ies rimarily n he

underlying thermal model, the Thermal Contrast Model 2-TCM2, with T A W S using a scaled-down version of that model. he SP M resident in TAWS uses the equivalent bar-chart approach but differs from Acquire

by irectly ncorporating lutter ffects or 0 ercent robability f acquisition. he T A W S predicts lock-on range based on signal-to-noise ratio hresholds, hot pot detection based n M D T methodology, nd discrimination etection ange based n M RT methodology. Thus, ll

other arameters being qual, he AW S etection anges hould be

approximately qual o hose redicted by Acquire t he 0 ercent

probability of detection level fo r moderate clutter (see discussion below).

Clutter is automatically computed, based on an empirical algorithm [14], and is a strong factor in determination of the number of cycles on target.

During the 980s, Dave Schmieder t the GTPJ examined he ffect of

clutter on target detection. 5,6] He found that the amount and nature of

background clutter had a significant impact on the probability of target

detection. chmieder first looked at scene radiance standard deviation as a clutter measure. owever, this measure had the deficiency of giving

large lutter values o elatively uncluttered cenes when hose cenes possessed several intensity modes. Moreover, this definition, like many other amplitude measures [5] , lacked a weighting factor based on target

size. oth amplitude and target size measures appeared to be required to

14



predict bserved rends. ince xisting efinitions ppeared

inappropriate, Schmieder redefined th e erm. he clutter definition he

used was a scene radiance standard deviation computed by averaging the radiance ariances f contiguous cene ells ver the whole cene nd

taking the square root of th e result. his is formulated as

clutter = [fVf INf1, 5)

i

where \ is the radiance standard deviation fo r the ith cell and N is the

number f ontiguous ells n he cene. his efinition mplicitly

included both target size nd ntensity measures nd produced higher

values fo r scenes that looked more complex and cluttered. t also avoided

yielding large lutter value or relatively uncluttered cenes that still

contained variations in intensity. dditionally, it accounted fo r clutter object izes lose o he arget ize weighing ore n he lutter calculation. However, as with other definitions of clutter, this definition

introduces ts wn et f roblems. cene magery s equired o

adequately determine o\ and eq (5) is not scale invariant but depends on the cell size elected fo r calculation. chmieder took the cell size to be square in shape with side dimensions of approximately twice the target

height. However, if the scene under examination contains more than one type of target (man, tank, bridge, etc.) the cell size must be redefined fo r

each target examined, resulting in different values fo r ci and N.

Based pon his efinition, chmieder erformed xperiments with

observers which howed that clutter could be categorized according to the ignal-to-clutter atio SCR) where he ignal was he emperature difference between the maximum/miitimum target temperature as seen

by ensor nd he background emperature where he lutter was defined as above. chmieder found that high clutter exhibited a SC R of

< 1, moderate clutter an SC R of 1 to 10 , low clutter an SC R > 10 and < 40 ,

and no clutter effects could be assumed if the SC R > 40 . urthermore,

Schmieder's data howed that fo r a detection probability of 50 percent, the number of cycles (line pairs) required fo r detection of a target varied

between 0.5 fo r low clutter to 2.5 fo r high clutter, with moderate clutter

requiring approximately 1.0 cycle.

Schmieder oncluded hat cquisition evels re trong unction f clutter s well s esolution nd hat ange rediction models must

include lutter ffects. ecause he umber f ine airs er arget-

angular ubtense ecessary or etection s nversely roportional o

detection ange, hanges n CR an be xpected o ignificantly alter

target-detection range predictions.

15



The TAWS detection methodologyuses Schmieder's SCR algorithm at the

50 ercent probability of detection evel o determine whether MRT or

MDT ormalization hould be used nd o alculate arget-acquisition ranges. he SCR may be hought f as he atio f hot/cold pot ATT

(i.e., maxmium/minimum arget emperature ackground ean

temperature) to clutter equivalent temperature (CT), thus

SCR = ATT/ CT- 6 )

The value of CT s determined by the IR scene complexity (a measure of

the umber f bjects n he cene ompeting with he arget) n

conjunction with he ackground cene ontrast Sc) n he arget's

vicinity. ince detection methodologyand acquisition range both depend

upon SC R (discussed in some detail below).

The AWS etermination f cquisition ange s ixed t 0 ercent

probability f detection. hus, s mentioned bove n he ection n

Acquire MRT methodology, a solution for the range can be found if the

target's ritical imension, arget nd ackground emperatures, nd

MRT sensor curve re ll known. hile TAWS uses this methodology,

there re ignificant ifferences rom Acquire—most otably n he

determination of the number of line pairs on target, N.

2.2.2.1 Resolved Versus Unresolved Targets

Whether he AWS omputes he etection ange ia MRT r MDT

methodology depends on whether the target is resolved or unresolved at

a given sensor to target range and the SCR at that range.

For resolved targets, if the SCR > 40, indicating little or no clutter, MDT is

used with he arget ontrast ased n he emperature ifference

between the hottest (or coldest) facet of the target seen by the sensor and

a given background; this is referred to as ATMAX. ven though the target

is esolved, Schmieder did not ecommend using MRT because he elt

that with uch high ignal o lutter > 0), t does not matter f the target is resolved or not, and MDT (hot/cold spot) detection would give a

longer detection range than MRT. f the SCR is < 40, MRT is used with

the target contrast, A T A V G , being the difference of the average temperature

of the target facets seen by the sensor and the background temperature.

16



If the target is unresolved at a given range, it is either detected with the

M D T methodology or is not detectable at all. f the unresolved target is

detectable using M D T , then SC R > 40 . n that case, ATMAX is used fo r the target and background contrast. f SC R < 40, and the target is unresolved, then the clutter is to o high, and no detection occurs at this range and the

range must e ecreased or etection o ccur. hese esults re

summarized in table 2.

Table 2. ummary of detection methods

SCR Line pairs on target Method

<1 2.5 M RT

< 1< 10 < 1.0 M RT

10<40 0.5 M RT

>40 - M D T

For alculating he CR n he bove rocedures, chmieder

recommended using the ignal ATMAX ather than ATAVG, fo r the reason

that in his 1982 SC R study [6] the signal component was defined as ATMAX

rather than ATAVG-

2.2.2.2 Determination of SC R

In the absence of actual imagery of the target area, Schmieder estimated

clutter from a combination of scene complexity and Sc. W e present his

methodology in some detail below.

Schmieder's SC R algorithmaffects the acquisition range by modifying the

number of line pairs (N) on target. At 50 percent probability of detection,

Schmieder's results can be formulated as [15]

N -- 1.64 H

arg

(log(50?) + 1.2)1M

R (7 )

17



fo r CR .1 . he A W S uses he arget height fo r the arget-critical

dimension, Hta r g. n order to determine SC R from eq (6 ) both the target

contrast ATT) nd he lutter emperature, Cr must be known. T, n turn, equires knowledge f both cene ontrast and cene omplexity.

Because the T A W S does not produce or use target- and background-scene

images, it cannot compute clutter as originally defined by Schmieder. o

rectify his roblem, chmieder 16 ] ecommended hat lutter e computed based n T . e ssumed hat n he verage, he arget

contrast, AT T, which represents the temperature difference between target

and background, would be qual to °C. chmieder then used eq (6),

with he verage TT C, oupled with CRs f 0 , 0, , nd ,

representing clutter states of none, low, moderate, and high respectively,

to etermine alues or r. chmieder tates 16] , Clearly, ther

assumptions f arget ignal nd CR would ead o ther ssociated values. he values shown (in table 3) ar e used as a starting point to be used until the results of further research can yield a better basis fo r these

choices.

A method fo r relating scene contrast (Sc) to the clutter levels is needed.

Schmieder hose 17] c anges f o , o , o , nd o correspond o he lutter evels f one, ow, moderate, nd igh, respectively. hese quantities (SCR, CT, Sc) and their relation to clutter

level are summarized in table 3.

Clutter Sc range level SCR CT degrees C) degrees C)

none 40 .0 5 <1.0

lo w 1 0 .2 > 1.0-£2.0

moderate 4 .5 >2.0-<4.0

high 1 2 >4.0

To bviate he eed or he ser o elect lutter evel, he A W S

implemented Schmieder's lgorithms to determine CT based on a user

input of IR-scene complexity and a calculation of scene-thermal contrast by the AW S IR model. 18] he lgorithm fo r determination of Sc as

found in TAWS, is presented in table 4. n this table, TB s a background

temperature nd he ubscripts nd 2, H and efer o he irst and

second, high nd ow background emperatures, espectively. able 5

and figure 3a show the original implementation.

18



Table 4. Algorithm for determining Sc Number of c value backgrounds degrees C) 1 .5

2 BI TB 2

3 BH - TBL

Table 5. nitial T A W S implementation of Or algorithm

Scene

Complexity

Sc degrees C)

Otol >lto2 >2to4 >4

r degrees C) None 05 05 05 05

L ow 05 2 2 2 Medium .05 .2 .5 .5

High .05 .2 .5 2.0

In operational use it was found that the step function shown in figure 3a

resulted in occasional discontinuities in detection range. o mitigate this,

the tep unction was eplaced with he ontinuous inear unctions

shown n able nd igure b. ence, n onjunction with cene

complexity, he arger he emperature ifference mong backgrounds

(i.e., the higher the scene contrast Sc), the higher the value of Or, which in

turn owers he alue f SCR o c Cr1)- he ower SCR s, he ewer

equivalent ine pairs here re per arget see q 7]), which ffectively

reduces the detection range. n general, as the number of backgrounds

increases he detection range will change, enerally decreasing in value

somewhat. his s easonable ince y dding ackgrounds, ne s

effectively increasing the clutter and making the target more difficult to

detect. hus, the lgorithm in TAWS for deterrriining Or is reasonable;

however, he boundaries or Or nd c, hosen by Schmieder 16,17]

using his experience and expertise, are subject to review. As indicated in

a previous paragraph or Or; nd s Schmieder pointed out 17 ] or Sc

These values result from heuristic and judgmental considerations. hey

have not been derived from sensitivity trades, which rigorously calculate

the lutter evels hat re btained rom arious background ontrast

conditions. uch omprehensive tudy will be ventually needed o

arrive at more fully supported values.

19



E

I 1 . 0

S c e n e Comple x i t y

— High

- -»Low

2 Sce ne Cont ras t Sc ( de gre e s )

2 S c e n e Cont ras t Sc ( de gre e s )

Figure 3a. nitial clutter temperature algorithm. Figure 3b. Current clutter temperature algorithm.

Table 6 . Current T A W S implementation of Or algorithm

Scene

complexity

Sc degrees C)

Otol >lto3 >3to4

CT degrees C)

>4

None

L ow

Medium

.05 .05

.05 +0.15 Sc .2

.05 +0.15 Sc

.05

.2

.5

.05

.2

.5

High .05 +0.4875 Sc 2.0

This clutter algorithm has an unexpected effect on MRT detection ranges.

When only one background is selected by the user (i.e., no scene-thermal

clutter), or if the temperature difference among the selected backgrounds

is < 1.0, the acquisition range found is the same whether the user selected

scene omplexity s ow r medium. ote he middle cene ontrast

categories now break at 3° rather than 2° for medium-scene complexity.

This change

is not significant since the

impact on clutter temperature

is at

most 15°, and he original distinction between low and moderate scene

contrast ategories was n rbitrary election etween he hreshold

values of 1 for none and 4 for high.

20



2.2.3 Discussion

Johnson's aper mentioned he ssue f ackground lutter. e

emphasized hat is etection riteria 1.0-line airs) were pplicable under onditions hat equired ome egree f arget hape

discrimination n rder o etect he arget rom ther bjects n he

background, (i.e., where background clutter was present). He also stated

that he umber f line airs equired o ttain a particular detection

probability ould ary ignificantly depending upon the ature f the

background lutter. his aised umber f uestions s o he

significance of background clutter on the validity of the Johnson criteria,

especially since Johnson did not provide a definition of clutter.

Schmieder noted that the Johnson criteria fo r target detection (1.0 cycles

or line pairs) correlated well with his moderate-clutter category (implying that uch evel f lutter was robably resent n ohnson's work, although ohnson id ot measure t). omparison f Schmieder's

detection probability as a function of resolution with Acquire's TTPF (see

figure 4) clearly shows that Acquire's formalization compares favorably with chmieder's oderate lutter. ther lutter evels an e

accommodated in Acquire by varying N50.

Figure 4. Probability

of detection versus

— Scluueder: L-

— — Schmierer: H

olution fo r 50 it acquisition

3 the T A W S 1 —| _ — ..... Acquire - r algorithms. a .« —

S -w ÄO i.4 —

0L , •

y

0. 1 —

i <s

1 z 3

Resolution Line Pairs / Target)

21



3.

Comparisons

3.1 cenarios

To ompare hese wo omplex arget-acquisition models equires

standardization of as many parameters as possible. o accomplish this,

one weather scenario was used in conjunction with one sensor and target,

both with fixed orientations.

A winter cenario was hosen nd xamined sing n xercised -80

Soviet main battle tank against two backgrounds (vegetation and snow)

at IR wavelengths. he sensor and tank were alignedsuch that the sensor always had a frontal view of the tank; the sensor height was fixed at 30 0 ft

facing orth. o minimize hadow ffects, he ate was ixed t 1 December at a local time of 12N. he location was also fixed at latitude of

37032' N/ longitude of 127°00' E (Seoul, S. Korea). he weatherconditions

include see able ) lear kies with arying isibility nd elative

humidity, nd vercast kies with arying isibility nd elative humidity. dditional ases ere un ncluding ight/heavy og

conditions, snow, drizzle, and rain.

Table 7 . Weather conditions used in the study

Relative Humidity

Precipitation

30 % none

50 % none

80 % none

100% light

fo g

100% heavy

fo g

80 % snow

90% light

rain

90% moderate

rain

Cloud clear clear clear clear clear

Cover overcast overcast overcast overcast overcast overcast overcast overcast

2km 2 2 2 2 2 2 2

Visibility 5km

10 km

5

10

5

10

5

10

rr; ---:-.- 5

10

5

10

15 km 15 15 15 \ 15 15

23



3 .2 Model Runs T he AW S was un using winter limatology along with the weather

conditions isted n able . omparisons were made with he cene

complexity nitially et t low. o educe he ossibility f rrors,

Acquire was initially run separately; however, this procedure was fraught

with problems (see table 8). o alleviate this, the Acquire algorithm was

programmed directly into TAWS , thereby, insuring that the many values

(AT, tmospheric ransmission) ere dentical n oth rograms.

Discussion onthe problem and their resolutionappear in table 8.

Table . roblems etween he AW S nd Acquire lgorithms nd heir

resolution

Problem Routine

TAWS Acquire

Resolution

Characteristic

dimension

Target height V(Xeff*Yeff) Target height

Sensor curves Horizontal (ID) Horizontal and

Vertical (2D)

ID

Aspect ratio V (7/2 Xeff/Yeff) <3 V (7/2 Xeff/Yeff)

Backgrounds 3 allowed N/A 1 used

Scene

complexity

None, low,

moderate, high

N/A+ Low

* ee section 3.2.>

3.2.1 Characteristic Dimension

The AW S nd Acquire se ifferent arget-characteristic imensions.

For M RT calculations, T A W S uses the target height; whereas, 2D Acquire uses the square root of the target's projected area as seen by the sensor. When Acquire was rogrammed nto he AWS , hereby using D formulization, the characteristic dimension was changed to use the target

height.

In th e course of this study the T A WS and Acquire target T-80 Soviet main

battle tank databases were examined; the results from T A W S and Acquire databases, nd dditional ources, re resented n able . he CASTFOREM is the U.S. Army's entity level warfare simulation; World

Wide W eb ( W W W ) 1, 2, and 3 were taken from various, unsubstantiated, W W W ources or omparison urposes. he alues elected or he

various imension izes robably epresent ifferent onfigurations f

24



the -8 0 cf. igures a nd ). sing he un-forward ength s not

representative of the actual target size and will produce overly optimistic

detection ranges. he TAWS database has been subsequently changed to reflect he more ccurate alues. nce Acquire was oded nto he

TAWS, both algorithms used the same database.

T-80B T-«CU

TT-V^v^i;.

Figure 5a. -80B Figure 5b. T-80U

Table 9. -8 0 dimensions

Source Length (m) Width m) Height m)

Acquire^ - 3.59 2.64

T A W S 9.1 4.64 3.73

CASTFOREM 6.75 3.55 1.5**

W W W 1 9 .7 3 .6 2.2

W W W 2 9.9/7.4§ 3 .4 2.2

W W W 3 7.01 3 .6 2.20

X values are fo r projected area

§ gun forward/hull

* * does not include turret

25



3.2.2 ensor Curves

Because he AW S was onstructed during he 980s, it uses D M R T curves, whereas, Acquire uses 2D M RT curves. nitially, before coding

the Acquire lgorithm nto AW S nd he wo odes were xecuted separately, it was necessary to ensure that th e two E R sensor curves that

were being used were the same. his was accomplished by multiplying

the TAWS abscissa by the normalization factor of .7 5 (see section 2.2.1);

the esulting omparison s presented n igure . s hown, the wo

sensor curves re dentical. hen Acquire was coded into T A WS and

thus used TAWS M RT curves, N50 fo r detection was changed from .75,

appropriate fo r th e 2D algorithm, to 1.0, appropriate fo r the ID algorithm.

Figure 6 .

Comparison of the

TAWS and Acquire

MRT curves.

GD I L — —

2£D lD W ID inn

3.2.3 Aspect Ratios

In ID Acquire, the target was described by its minimuin dimension; thus, targets with dimensions (length x width) of2mx2mor2mx4mor2mxl6m

and identical AT would al l be equally detectable according to ID Acquire. The aspect ratio adjustment compensates fo r this unrealistic result. his

is an important issue fo r large aspect ratio targets such as battleships but

not fo r th e ypical aspect ratios of ground vehicles. n 2D Acquire, the

target is described by the quare oot of its area; thus, the 2 m x 6 m

26



3.2.4 ackgrounds

target will have a higherestimated probability of detection than the 2 m x

2 m target. herefore, in moving from ID to 2D Acquire, we have gone

from th e simplifying assumption that al l targets of the am e height ar e equally detectable (given th e same AT) to the simplifying assumption that

all targets with the same presented area are equally detectable; however,

aspect ratios > 3 are not recommended fo r Acquire. [3 ] or the typical

aspect atios f round ehicles, his s easonable implifying

assumption fo r 2D Acquire.

When sing D cquisition models uch s AWS , he irst tep conventionally performed in range prediction is to convert the laboratory

M RT o ak e nto ccount he cene bject spect atio maximum o

minimum object dimension) if use is to be made of the concept that an object is more readily discerned if the aspect ratio is greater than unity. The laboratory M RT is computed or measured with a bar aspect ratio of 7.

It can be shown [19] that

MRTfieid MRTj7/ 2Ne), 8)

where s is the scene aspect ratio.

For xample, or etection nly ne ycle s equired N 50 ),

yielding MRTfieid = MKlJl/2e Within TAWS , the M RT is adjusted

by a factor oi lXeff /2Yeff where Xe f f (along-track) and Yef f (cross-track)

are the projected target dimensions at a given range. he cross-track and along-track dimensions, viewed in a plane coincident with the sensor, ar e the bscissa nd rdinate, respectively. ince he alculation of aspect

ratio s ntegral o AW S alculations, o hanges were made o he

algorithm.

The AW S l lows alculations or cenes hat nclude p o hree backgrounds; whereas, Acquire does notconsider differing backgrounds,

primarily because Acquire ccounts fo r clutter through variation of the

parameter N50. he TAWS backgrounds ar e intimately connected with

the clutter calculations (see section 2.2.2), allowing ATT in eq (6 ) to vary

as the backgrounds ar e cycled. hus, eq (6 ) can be rewritten as

27



SCR AT / CT (9 ) 7= 1

3.2.5 Clutter

Figure 7 . Acquire versus the TAWS clutter comparison.

where he ubscript efers o he ackground nder onsideration

(considered primary) nd he ubscript; efers to the number of (user- entered) backgrounds NB maximum of 3) . hus, the T A W S detection

ranges fo r all backgrounds ar e calculated with the displayed background,

as printed on the output, considered primary. discussion of the lack of

backgrounds n Acquire, which re ntimately ied o he lutter

calculations in TAWS, is deferred to the clutter section below.

Schmieder's work, as implemented in TAWS, accounts fo r various levels

of lutter; whereas, Acquire nly ccounts or lutter by arying N50.

However, s hown n ection , Acquire's lgorithm, with 50

(detection) ompares avorably with Schmieder's t moderate lutter levels. o effectively represent Schmieder's low and high clutter cases, N50 takes on values of 0.5, and 2.5 respectively (see figure 7) . ince clutter is a subjective measure, othervalues of N50 may be chosen.

A— Schjnieder: L«w

Schmied er: Moderate Sclwieder: High Ac««xre,N50 = 1 .0 Ac«ni», N S O = O. S Acquire, NSO = 2.5

Resolution Line Pairs / Target)

28



3.3 Results

The December 1

ases were un sing ow-scene omplexity

with

single vegetation background in TAWS, and with N50 = 0.5 in Acquire. Results showing the detection ranges predicted at noon in each case by

TAWS andAcquire ar e shown in figure 8.

Examination f he ata hows maximum alue or he A W S

calculations t 2.1 m or he egetation background ue o ensor

optical resolution limits. ecause this limits scales with target height, a

determination an e ade sing alues aken rom eference

18 tables A.l-2. or the sensor and target chosen th e range limit turns out

to be 32.8 km in good agreement with the TAWS values. his same cutoff

fo r a maximum detection range was applied to the Acquire results.

For detection ranges less than 10 km the agreement between Acquire and TAWS was also good. etween these high and lo w detection ranges, i.e.,

midrange alues etween 0 nd 0 m, cquire redictions were

10 to 15 percent lower than the TAWS. pecifically, the cases with heavy

fo g or precipitation have identical values below 10 km, while the cases

with lo w or moderate humidity and high visibilities have values of km. Thus, e ay raw he onclusion hat he ontribution rom

approximately 2 lutter s rrelevant n nstances when he weather conditions ar e either extremely unfavorable (low visibility) or extremely

favorable (high visibility).

Figure 8 . he S versus

lire-detection iges at 5 0 percent

probability of detection.

1 0 0 0 0 TAWS Detection Range (km)

29



Some of the discrepancy between the T A W S and Acquire in the midrange

may be explained by the fact that the clutter temperature in the T A WS

runs with one background and low-scene complexity is 0.12S °C (tables 4 and 6 ), which equates to an N50 value of approximately .3 , while the value

used or hese Acquire uns N 5o .5 ) ould be ssociated with

low-clutter alue f 0.2 C. owever, note hat Acquire does not use

clutter temperature directly. ncreasing the clutter temperature to 0.2 °C in the TAWS, by including an additional background with an appropriate

scene ontrast emperature, id ower he AW S ange ut nly

accounted fo r less than half of the discrepancy between the T A W S and

Acquire values. When additional Acquire runs were made with N50 se t

= .3 , which is not strictly analogous to calculating a clutter level in TAWS,

the results were within 5 percent of the T A W S values.

3.3.1 lutter/Complexity Effects

Because he ser an elect cene omplexity evel f one, ow,

moderate, or high in the T A WS (specified as Clutter/Complexity in the

User Interface), it is important to understand how this selection affects the

resulting etection ange. nly ne background was sed or hese comparisons in order to hold constant th e cene contrast component of

clutter. hese omparisons howing he ffects f the hoice f scene

complexity n A W S nclude he omplete ourly ata rom he

December 1 uns, ather han just he oontime ata. he tandard

deviations of the difference between T A W S and Acquire detection ranges throughout each daily un re mall, ndicating that diurnal variations

behave similarly in both T A W S and Acquire. n general, most cases using

one background displayed a decrease in detection range of approximately

5 ercent when cene omplexity was ncreased rom none o ow r

moderate, and nother 5 ercent ecrease when scene omplexity was

increased from low or moderate to high. However, there is a fair amount

of ariation rom his ypical esult, s hown by he ample f runs

plotted in figure 9. n this figure, the abscissa represents the cases run

(generally in order of cases with low humidity on the left to cases with

high humidity and precipitation on the right), and the ordinate showing

the percent of the resulting detection range with

a given

level of scene complexity compared o he ange ound with scene omplexity et to

none. Note from tables 4 and 6 that with only one background low- and

moderate-scene complexity will return the same values.

30



Figure 9 . cene complexity effects on target- detection ranges.

Low/Moderate High ComplexH

86.0

C a s e s with I n c r eas ing Hum id i t y /P r ec ip i t a t i on

The primary exception to th e general impact of varying scene complexity

occurs both when TAWS detection ranges ar e very long and when they

are ery hort. or xample, esults lotted oward he ef t ide f

figure 9 include cases where humidity is 50 percent or less, and there is

no fo g or precipitation. hese conditions result in long detection ranges,

which ar e not substantially decreased when the scene complexity level is

increased. dditionally, ases lotted n he ight ide f he raph provide very short detection ranges due to fo g or precipitation; so that

increasing

he

omplexity

evel

as

o

orresponding

ecrease

n detection range, since the detection range is already so restricted. Other

deviations re elated o he elected isibility n ach ase, with increasing cene omplexity ausing mpacts more han he sual

5 percent when visibility is great. In order to highlight the general trend,

cases with visibilities of 2 and 5 km have been omitted from figure 9 , which would otherwise show even greater fluctuations. he cases potted with visibilities of 5 and 10 km reflect some fluctuation relatedto whether

th e cloud cover was entered as clear or overcast, which accounts fo r the

up-and-down nature of th e graph.

Another esult ccurs when multiple backgrounds re elected n he

TAWS. he previous results were based on cases using vegetation as the only background. f econd ackground s dded, t ffects he

detection range predicted fo r the first background if the complexity level is et o nything ther han one. he ffect f dding econd

background s hown n able 0, sing he oontime alues nd

excluding the cases reflecting extreme high or lo w detection ranges. A s

31



expected, detection ranges fo r the vegetation background decrease when

the econd background f now s dded, ince he lutter has been

increased. Additionally, detection ranges fo r the vegetation background increase when now s ntered s he rimary ackground with vegetation s the econdary background. his is expected because the

thermal contrast (AT) betweenthe target and the background is calculated

using the first, or primary, background entered in TAWS; the secondary

background does not interact with the target. ecause snow has a higher

albedo han he egetation, elatively reater mount f he olar

radiation is reflected onto the tank surface raising its temperature; and,

thus, producing the larger contrast value.

Table 10. mpact of a second background on detection range

TAWS Range km)

Scene

complexity

Vegetation

background

only

Vegetation Vegetation

background background

primary with secondary with

snow background snow background

secondary primary

None 22.46 22.46 23.30

L ow 21.46 21.36 22.22

Moderate 21.46 21.07 21.95

High 20.39 19.87 20.81

A s discussed in section 2.2.2, rather than a second choice of a background

fo r a what if capability to see two separate background results in a single model un, dding econd background erves o dd lutter o he

scene, with he reatest mpact n onjunction with he election f moderate-or-high complexity. Note that when multiple backgrounds are selected, there ca n be a difference between detection ranges based on low and moderate omplexity, s well s ven reater ifferences between moderate and high complexity than seen with just one background. his is easonable result, and has been highlighted in the updated T A W S user ocumentation Version .1 nd reater). n cases with multiple

backgrounds, the clutter temperature Or will not necessarily be consistent fo r each time in a 24-hour run, or between cases with varying weather

input, ue o he ifferential eating f he arget nd he ifferent

background types. However, to get an idea of the total effects possible

under different clutter amounts, a single example at a single time shows a

32



decrease in detection range from 19.69 km based on a single background

and no cene omplexity to 5.75 km based on three backgrounds and

high-scene complexity.

3.3.2.2 Weather Effects

Although one purpose of this report is to compare the predicted target-

detection anges utput by he AW S nd Acquire, t s lso worth examining he ffects f weather n he redicted anges. arying

visibility, relative humidity, cloud cover, fog, and precipitation resulted

in imilar impacts o etection ranges in both the T A W S and Acquire,

consistent with at least qualitative expectations of how these atmospheric

properties affect IR sensors. [20,21]

The

ollowing

xamples

re

ased

n

airly

ealistic

winter-weather scenarios fo r Seoul, Korea. limatological temperature values varied 6 °C over he 4-hr eriod, with ess ariation n ew-point emperature

values, resulting in a typical increase in relative humidity values in the

early morning and lower relative humidity in the afternoon. hese same temperature values were used whether or not clouds, fog, or precipitation

were included. hese examples ar e based on low-scene complexity and

one background in the TAWS, and N5o = 0.5 in Acquire. s discussed above, ncreasing he omplexity evel o igh rovided omewhat shorter detection ranges in most cases, but yielded similar results in terms

of weather mpacts n etection-range redictions. able 1 ists he noontime detection ranges calculated by the TAWS (number in upper left

of ell) nd Acquire number n ower ight f cell) or ac h weather scenario. s previously discussed, the Acquire values ar e not resolution limited to th e T A WS' appropriate maximum of 32 km in cases with high

visibility and no fo g or precipitation. Other cases, which include fo g or precipitation seem to return detection ranges around 7 km using Acquire, while he AW S rovides much ower etection anges. therwise,

Acquire esults how imilar mpacts based n weather nd iurnal

effects s he AWS, nd ubsequent iscussions il l ighlight he

specific-weather impacts using the T A W S data.

33



Table 11 . Detection ranges as a function of various weather scenarios

Detection Ranges (km) TAWS

Acquire

RH Precipitation

30% none

50% none

80% none

10 0 % light 100% heavy

80% snow

90% light rain

90% moderate

rain

Cloud Cover

S - 4 - > cß c d u 1 1 0

> o

- 4 - » e n c d C D > O

c

4- * c d Ö > o

•S i c d u >

3 ovec

le r

ovec

le r

ovec

le r

ovec

? 2 1 1 8 2 0 1 8 1 9 1 6 5 5 2 2 2 5 3

1 « 1 6 1 7 1 5 1 6 1 4 5 5 2 2 2 5 S 5 3 2 2 8 3 0 2 7 2 8 2 4 1 2 1 0 4 5

? 9 7 5 2 7 3 2 2 3 2 4 2 0 1 0 9 4 5

1 0 3 2 3 2

3 1 3 2 2 8 1 8 1 5 7 6

• 1 H 3 2 3 2 3 2 2 9 3 0 2 5 1 5 1 4 7 6 tß > 1 5 3 2 3 2 3 2 3 2 3 2 2 9 2 1 1 8 1 0 6

6 3 2 3 2 3 2 3 2 3 2 2 7 1 8 1 6 9

3.3.2.1 Visibility

Although R ensors re seful or etecting argets t anges

beyond the distance visible to the unaided eye, he same

atmospheric onstituents, which educe isibility, will reduce R

detection anges, lthough o ifferent mount. Figure 0

illustrates he mpact of reduced visibility on TAWS. Under the

benign conditions of low-relative humidity (around 30 percent at

noon), clear skies, and 15 km isibility, TAWS provides a

maximum ensor-detection range imit of 32 km throughout the

24-hour period. The other examples with clear skies result in an

increase n etection anges during aylight ours, s olar

loading heats the target more than the background, resulting in a

greater T or ifferential n arget/background emperatures).

Visibility decreasing from 15 to 10 km generates rmnimal impacts,

but oing rom 0 o m auses 5 ercent ecrease n

detection range, while going from 5 to 2 km results in a 35 percent

decrease.

34



Figure 10. Visibility impacts on the TAWS- detection ranges.

— — — — -— — . ? <_- ^—, D E X ^v S ? - a c o

yf ^ V > yT N^_

^^f — —

DC — 15

10

R H = Low Clear Skies

isby=15km isby=10km isby=5km isby=2km

ID 11 12 13 14 15 16 Local Time IB 20 21 22 23 54

3.3.2.2 elative Humidity

Atmospheric moisture absorbs IR signals. ecause the time series

of emperature alues s onsistent n ac h ase, hanging he

relative humidity is equivalent to changing the amount of water vapor or absolute humidity available fo r attenuation of the sensor

signal. igure 1 hows n xample f ncreasing umidity

resulting n ecreasing etection anges. ince he old winter

temperatures used in these cases do not allow the atmosphere to contain ignificantly more moisture, the etection range s only

decreased y 0 ercent s umidity s aried rom ow

(approximately

0

ercent;

quivalent

o

/m

3

)

o

igh (approximately 80 percent; quivalent o /m3). owever,

comparable runs made using summer temperatures show noon-

time etection ange redictions f 8 m n ow umidity

(approximately 30 percent; equivalent to 7 g/m3) falling to 13 km

in igh umidity approximately 0 ercent; quivalent o

19 g/m3), reflecting more than a 50 percentdecrease.

35



C le a r S k i e s Visby=5km Figure 11. Relative humidity impacts on the

TAWS-detection ranges.

35

— 0 i 0)

g>25 is a

1 3 0 a a ; a 15

10

RH=Lo w (30 a t n o o n )

H = M o d e r a t e ( 50 at n o o n )

H=High ( 80 a t n o o n )

2 3 4 5 6 7 8 9 2 3 4 Lo c a l T i m e

3.3.2.3 ky Cover

The primary effect of cloud cover above the sensor path is based on the

cloud's nfluence n eating nd ooling f he arget nd he

background. s shown in figure 12 , this example exhibits slightly longer

detection ranges during the night with overcast skies compared to clear

skies. lthough both the arget and the background temperatures are

affected by he louds, o hat he T emains maller han when o clouds ar e present, less radiational cooling allows the thermal imager to

detect the warmer target at slightly longer ranges than under clear skies.

Cloud cover has a reater impact during the day, as the reduced solar

loading provides only a 5 percent detection range increase, while the case with no clouds produces a twenty percent greater detection range during

the daytime.

36





Figure 13 . og

impacts on the

T A W S - detection

ranges.

RH=High Overcast Visby=2km

-LIGHT FO G (Radiat ion)

—HEAVY FO G (Advect ion)

9 0 1 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 2 1 2 2 2 3 24 Local Time

RH=High Overcast isby=5kn[ Figure 14 . Precipitation

impacts on the TAWS-detection

ranges.

LIGHT RAIN (2 m m / h r )

-MODERATE RAIN (5 m m / h r )

- S N OW

2 0 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 2 2 1 22 23 24 L o c a l Time

8



4 . onclusions

It is important that Acquire be merged into the T A W S so that the services

ca n redict arget cquisition f ground argets using ecognition nd

identification in addition to detection. he incorporation of the Acquire

SP M into the T A W S is scheduled fo r the TAWS Version 3, which will be

released or se n 001. hese omparisons f the arget-acquisition

range output from the urrent version of the T A W S with output from

Acquire provide positive feedback on the benefits of this enhancement to

the AWS , while maintaining he xisting nterfaces nd roviding

comparable arget-detection range predictions.

he primary benefit

of this nhancement il l e he bility o pecify arget-acquisition

discrimination levels, including detection, recognition, and identification.

Decisions will need to be made on an efficient and appropriate selection

of N5o fo r use in the Acquire SPM.

Because the cases examined in this report ar e limited to a winter scenario

in Korea, specific quantitative results of the selected weather parameters'

impacts on target-detection range cannot be generalized to al l situations.

However, these cases do highlight the importance of considering accurate

atmospheric onditions n arget-acquisition redictions. esults

showing maller weather mpact o Acquire-detection anges han

predicted sing he urrent AW S PM nder onditions f og r precipitation warrant additional investigation.

39



References

1 . auter, D., M . orres, .D . Brandt, t al., The ntegrated Weather Effects Decision Aid: A Common Software Tool to Assist in Command

and Control Decision Making/' roceedings f the Command Control Research Technology Symposium, Newport, RI, (June 1999).

2. ouveia M.J., J.S. Morrison, R.B. Bensinger, et al., T A W S and NOWS:

Software Products fo r Operational Weather Support, roceedings of the Battlespace Atmospheric an d Cloud Impacts on Military Operations Conference, Colorado State University, Fort Collins, Colorado, (25-27 April 2000).

3. Acquire Range Performance Model for Target Acquisition Systems, Version 1

User's Guide, U.S. Army CE COM Night Vision and Electronic Sensors

Directorate Report, Ft. Belvoir, VA , (1995).

4. ohnson, J., Analysis of Image Forming Systems, Proceedings of the Image ntensifier ymposium, .S. rmy ngineer esearch nd Development Laboratory Ft. Belvoir, (A D 22 0 160), (October 1958).

5. chmieder, D.E., and M.R. Weathersby, Detection Performance in Clutter with ariable Resolution, IEEE Transactions on Aerospace and Electronic

Systems, Version 19 , pp. 622-630, (July 1983).

6 . chmieder, D.E., M.R. Weathersby, W.M. inlay, t al., lutter an d Resolution ffects n Observer Static Detection Performance, U.S. Air Force

Wright Aeronautical Laboratory, echnical Report AFWAL-TR-82-1059 (A D B071777), Wright-Patterson AFB, OH, (June 1982).

7. ouart, C.N., M.J. Gouveia, DA. DeBenedictis, et al., Electro-Optical Tactical ecision id EOTDA) ser's anual, ersion , echnical Description, U.S. Air Force Phillips Laboratory, Technical Report PL-TR-

93-2002 (A D B172088L), Hanscom AFB, M A , (June 1994).

8. Army Modeling an d Simulation Office, Standards Category Acquire, web site, http://www.amso.army.mil/.

9. ohnson, J., and W.R. Lawson, Performance Modeling Methods and

Problems, Proceedings of the IRIS Specialty Group on Imaging, pp.

105-123, Infrared nformation nd Analysis enter, RIM, Ann Arbor, M I,

(January 1974).

10 . Mazz, J., Acquire Model: Variability in N50 Analysis, Proceedings 9th

Annual round arget odeling nd alidation onference, ignature

Research, Inc., Calumet, M I, (August 1998)

41



11. Howe, J.D., Electro-Optical Imaging System Performance Prediction, in The Infrared an d Electro-Optical Systems Handbook (U), volume4., M .C. Dudzik, Ed, Infrared Information Analysis Center and SPIE Optical Engineering

Press, (1993).

12. erk, ., .S. Bernstein nd . . obertson, MODTRAN: Moderate Resolution Model for LOWTRAN , U.S. Air orce Geophysics

Laboratory, echnical Report, GL-TR-89-0122 AD A214337), Hanscom

AFB, M A , (1989).

13. hirkey, R.C., L.D. Duncan and F.E. Niles, The Electro-Optical Systems Atmospheric ffects ibrary, Executive Summary, Atmospheric ciences

Laboratory, echnical eport, SL-TR-0221-1, White ands issile

Range, NM, (October 1987).

14. iggins, G.J., .F . Hilton, . hapiro, t l. , perational actical Decision Aids (OTDA)s U.S. Air Force Geophysical Laboratory, Technical

Report, GL-TR-89-0095 (A D B145 289), Hanscom AFB, M A , (March 1989).

15. Higgins, G., D.A. DeBenedictis, M.J. Gouveia, t al., lectro-Optical Tactical Decision A id EOTDA ) Final Report, U.S. Air orce Geophysics Laboratory, echnical eport, L-TR-90-0251 I) AD 153311L),

Hanscom AFB, M A , (September 1990).

16 . chmieder, D.E., echnique for ncorporating High Resolution arget Signature Predictions into Sensor Performance Models, U.S. A ir Force Wright Aeronautical Laboratory, Technical Report, AFWAL-TR- 87-1055, Wright-

Patterson AFB, OH, (July 1987).

17.

chmieder, D.E., Interim Contrast Categories for 'Rule of

Thumb' Clutter

Computation, Georgia Technical Research Institute, Technical Transmittal,

TT-4828-004, Atlanta, GA , (March 1988).

18. ouart, C.N., M.J. Gouveia, D.A. DeBenedictis, et al., Electro-Optical Tactical ecision id EOTDA ) ser's anual, ersion , echnical Description, ppendix , hillips aboratory echnical eport PL-TR-93-2002 (II) (A D B171600L), Hanscom AFB, MA, (January 1993).

19 . oist, .C., lectro-Optical maging ystem erformance, CD

Publishing, Winter Park, FL , (1995).

20 . uantitative Description of Obscuration Factors for lectro-Optical an d Millimeter ave ystems, ilitary andbook, OD-HDBK-178(ER),

(July 1986).

21 . Federation of American Scientists, METOC Effects Smart Book (U), web site, http: / /www.fas.org/spp /military/program/met/metocsmarttbook.htm

42



Acronyms

ID 2D

A MSA A

AFRL

CASTFOREM

EOTDA

ERDL

FLIR

GTRI

IFOV

IWEDA

IR

M DT

M RC

M RT

NVESD

SC R

SPM

TASC

TAWS

T CM2

TDA

TRAC

one dimensional two dimensional

Army Materiel Systems Analysis Agency

U.S. A ir Force Research Laboratory

Combined Arms and Support Taskforce Evaluation

Model

Electro-Optical Tactical Decision Aid

U.S. Army ngineer esearch nd Development

Laboratories

forward-looking infrared

Georgia Tech Research Institute

instantaneous field of view

Integrated Weather Effects Decision Aid

infrared

minimum detectable temperature

inimmum resolvable contrast

minimum resolvable temperature

Night Vision and Electronic Sensors Directorate

signal-to-clutter ratio

Sensor Performance Model

The Analytic Sciences Corporation

Target Acquisition Weather Software

Target Contrast Model 2

tactical decision aid

TRADOC Analysis Center

43



T R A DOC raining and Doctrine Command

TTPF arget Transform Probability Function

W W W orld Wide W eb

44



Distribution

Copies NASA MARSHALL SPACE FL T CTR

ATMOSPHERIC SCIENCES DIV

ATTN DR FICHTL HUNTSVILLE AL 35802

NASA SPACE FL T CTRATMOSPHERIC SCIENCES DIV

CODE ED 411 HUNTSVILLE AL 35812

USARMYSTRATDEFNSCMND

CSSD SL L ATTN DR LILLY

PO BOX 1500 HUNTSVILLE A L 35807-3801

US ARMY MISSILE C M N DA M S M I R D AC A D ATTN DR PETERSON REDSTONE ARSENAL AL 35898-5242

US ARMY MISSILE C M N D

A M S M I RD AS SS ATTN MR H F ANDERSON REDSTONE ARSENAL A L 35898-5253

US ARMY MISSILE C M N DA M S M I RD AS SS ATTN MR B WILLIAMS REDSTONE ARSENAL AL 35898-5253

US ARMY MISSILE C M N DA M S M I RD DE SE ATTN M R GORDON LILL JR REDSTONE ARSENAL A L 35898-5245


REDSTONE SC I INFO CTR A M S M I RD CS R DOC REDSTONE ARSENAL AL 35898-5241

US ARMY MISSILE C M N DA M S M I REDSTONE ARSENAL AL 35898-5253

45



PACIFIC MISSILE TEST CTR GEOPHYSICS DIV ATTN CODE 3250

POINT M U G U C A 93042-5000

N A V A L O C E A N SYST CTR

CODE 54 ATTN DR RICHTER SA N DIEGO CA 52152-5000

METEOROLOGIST IN CHARGE KWAJALEIN MISSILE RANGE PO BOX 6 7 A PO SA N FRANCISCO CA 96555

DEPT OF C O M M E R C E CTR

M O UN T A I N ADMINISTRATION

SPPRT CTR LIBRARY R 51

32 5 S BROADWAY BOULDER C O 80303

N C A R LIBRARY SERIALS NATL CTR FO R ATMOS RSCH PO BOX 3000 BOULDER CO 80307-3000

DAMIPOI WASHINGTON DC 20310-1067

LEANDER PAGE D A M I POB WASHFATGTON DC 20310-1067

M IL ASST FO R ENV SC I OF C OF THE UNDERSEC OF DEFNS FO R RSCH & ENGR R & A T E LS PENTAGON ROOM 3D129 WASHESTGTON DC 20301-3080

US ARMY INFANTRY ATSH CD CS OR ATTN DR E DUTOH FT BENNPMG GA 30905-5090

A IR WEATHER SERVICE TECH LIBRARY FL4414 3 SCOTT AFB IL 62225-5458

USAFETAC DNE ATTN M R GLAUBER SCOTT AFB IL 62225-5008

46



HQ AFWA/DNX

10 6 PEACEKEEPER DR STE 2N 3 OFFUTT AFB NE 68113-4039

PAULTATTELMAN AFRLVSBE

29 RANDOLPH ROAD HANSCOM AFB M A 01731-3010

DRANDYGOROCH N A V A L RESEARCH LABORATORY

MARINE METEOROLOGY DIV CODE 7543 MONTEREY C A 3943-5006

MELANIE GOUVEIA

TASC 55 WALKERS BROOK D R READING M A 1867

US ARMY

MATERIEL

SYST ANALYSIS ACTIVITY

AMXSY ATTNMRJMAZZ A PG M D 21005-5071

US ARMY MATERIEL SYST

ANALYSIS ACTIVITY AMXSY AT ATTN M S R KISTNER APG M D 21005-5071

US ARMY MATERIEL SYST ANALYSIS ACnVITY

AMSXY APG M D 21005-5071

US ARMY RESEARCH LABORATORY AMSRLD 2800 POWDER MILL ROAD ADELPHI MD 20783-1145

US ARMY RESEARCH LABORATORY AMSRL O P CI SD TL 2800 POWDER MILL ROAD ADELPHI MD 20783-1145

US ARMY RESEARCH LABORATORY

AMSRL SSSH ATTNDRSZTANKAY 2800 POWDER MILL ROAD ADELPHI M D 20783-1145

47



US A R M Y RESEARCH LABORATORY

AMSRLIS ATTNJGANTT 2800 POWDER MILL ROAD ADELPHI MD 20783-1197

US A R M Y RESEARCH LABORATORY AMSRL 2800 POWDER MILL ROAD ADLEPHI M D 20783-1145

NATIONAL SECURITY AGCY W 21 ATTN D R LONGBOTHUM 9800 SAVAGE ROAD FT GEORGE G M E A D E M D 20755-6000

ARMY RSRC OF C

ATTN AMXRO GS DR BA C H PO BOX 12211

RTP NC 27009

DR JERRY DAVIS NCSU PO BOX 8208 RALEIGH NC 27650-8208

USARMYCECRL CECRL GP ATTNDRDETSCH HANOVER NH 03755-1290

US A R M Y A R D E C SMCARIMIIBLDG59

DOVER N J 07806-5000

US ARMY COMMUNICATIONS

ELECTR CTR FOR EW RSTA AMSRL EW D FT M O N M O UT H NJ 07703-5303

US ARMY COMMUNICATIONS

ELECTR CTR FOR EW RSTA AMSRL EWMD FT M O N M O UT H NJ 07703-5303

US ARMY DUGWAY PROVING GR D STEDPMTDAL3

DUGWAY UT 84022-5000

US ARMY D U G W A Y PROVING GR D

STEDPMTM ATTN M R BOWERS DUGWAY UT 84022-5000

48



DEPT OF THE AIR FORCE OL A 2D WEATHER SQUAD MAC H O L L O M A N AF B NM 88330-5000

PL W E KBRTLAND AF B NM 87118-6008

USAF ROME LA B TECH

CORRIDOR W STE 26 2 RL SUL 26 ELECTR PKWY BL D 10 6 GRIFFISS AFB N Y 13441-4514

AFMC D OW WRIGHT PATTERSON AF B OH 45433-5000

US ARMY FIELD ARTILLERY SCHOOL ATSFTSMTA FT SILL O K 73503-5600

US ARMY FOREIGN SC I TECH CTR

C M 220 7TH STREET NE CHARLOTTESVILLE V A 22448-5000

N A V A L SURFACE WEAPONS CTR

CODE G6 3 DAHLGREN VA 22448-5000

US ARMY OE C CSTE EF S PARK CENTER IV 4501 FORD AVE ALEXANDRIA VA 22302-1458

US ARMY CORPS OF ENGRS ENGR TOPOGRAPHICS LA B

ETLGSLB FT BELVOIRVA 22060

US ARMY TOPO ENGR CTR

CETEC ZC 1 FT BELVOIR VA 22060-5546

US ARMY NUCLEAR CM L AGCY M O N A ZBBLDG 2073 SPPJNGFEELD VA 22150-3198

US ATRADOC A T C D F A FT M O N R O E VA 23651-5170

49



US ARMY TRADOC ANALYSIS CTR

ATRC W SS R WSMRNM 88002-5502

DTIC 8725 JOHN J K I N G M A N RD

STE0944 FT BELVOIR VA 22060-6218


A M S M I REDSTONE ARSENAL AL 35898-5243 US ARMY DUGWAY PROVING GR D

STEDP3 DUGWAY UT 84022-5000

W S M R TECH LIBRARY BR

STEWS M IT WSMRNM 88002

US MILHARY ACADEMY DEPT OF MATHEM ATICAL SCIENCES

ATTN M D N A MAJ DON ENGEN

THAYER HALL WEST POINT NY 10996-1786

ARMY MODELDvTG & SIMULATION OFFICE

DDCSOPS ATTN D A M O ZS 40 0 ARMY PENTAGON WASHINGTON DC 20310-0450

US A R M Y RESEARCH LABORATORY

AMSRLCIEW ATTNDRSHIRKEYINFO SC I & TECH DIR WSMRNM 8002-5501

US ARMY RESEARCH LABORATORY

AMSRLCIEW ATTNBSAUTER INFO SC I & TECH D IR WSMRNM 8002-5501

M R RENE CORMIER AFRLVSBLDRC

29 RANDOLPH ROAD HANSCOM AF B M A 01731-3010

50



MRDDIXON

TRAC ATTN ATRC W BC W S M R N M 8002

Record Copy

95 TOTAL

Date post:	04-Jun-2018
Category:	Documents
Upload:	sapane
View:	221 times
Download:	0 times

Weather Effects on Target Acquisition

Documents