DTIC ADA081324: Field Manual to Determine Detection or Recognition Range of a FLIR Sensor.

Survival, Water, Medical Field Manuals

Military Manuals

Defense Technical Information Center

Document text

AD-A081  324  INSTITUTE  FOR  DEFENSE  ANALYSES  ARLINGTON  VA  SCIENCE  A— ETC  F/G  17/5 

FIELD  MANUAL  TO  DETERMINE  DETECTION  OR  RECOGNITION  RANGE  OF  A  F— ETC(U) 
SEP  79  L  N  SEEKAMP  MDA-903-79-C-0202 

UNCLASSIFIED  IDA-P-1419  S0IE-AD-E5OO  112  NL 


Ql ~>£S#o  //X 

Copy  16  200  copies 


fS)  level 


y 

ffiA  PAPER  P-1419 


FIELD  MANUAL  TO  DETERMINE  DETECTION 
OR  RECOGNITION  RANGE  OF  A  FLIR  SENSOR 


Lynne  N.  Seekamp 


September  1979 


DISTRIBUTION  STATEMENT  A 

Approved  lor  public  release; 
Distribution  Unlimited 


DT1C 

SELECTED 
MAR  5  19803  ■ 


Prepared  for 

Office  of  the  Under  Secretary  of  Defense  for  Research  and  Engineering 


1  1  0j0 


INSTITUTE  FOR  DEFENSE  ANALYSES 
SCIENCE  AND  TECHNOLOGY  DIVISION 


IDA  Log  No.  HQ  79-21390 


The  work  reported  ki  (Ms  document  was  conducted  under  contract 
MDA  903  79  C  0202  tor  the  DetMftment  of  Defense.  The  pubdeathn  of 
tlHs  DA  Paper  does  not  Indicate  endorsement  by  the  Department  of 
Defense,  nor  should  the  contents  be  construed  as  reflecting  the  official 
position  of  that  agency. 


Approved  hr  pubic  rohaso;  dhtrhuthn  unlmtted. 


UNCLASSIFIED 


tfCUMITy  ClA«H»ICATIQN  Q»  T»i«  »»0C  'On  D*t  tm.r.t, 

P””*  REPORT  DOCUMENTATION  PAGE 


I  READ  INSTRUCTIONS 

I  BEFORE  COMPLETING  FORM 


■  C»OAT  MuUMi 


il.  SOVT  ACC  11410*  no  |  1  ■CCi»>!C(«  r-J  CarNLOi:  <UNICI> 


*■  TiTl*  JuMtjIA) 

J  )  J1 ELD  ilANUAL  TO  DETERMINE  .DETECTION  OR 
[pt  TSeOGnSFION  RANGE' OF  A  FLlJTSENSOR  , 


Tvme  or  me/>o*T  o  rcmao  cpve»e o 


1  FINAL 


h-e  pT>± 


H  >UThON<; 


5U  Lynne  N.jseekanp 


(  MWroiSilfftfSsnreSoMT  mumrc*- 

IDA  PAPER  P-1419 _ 

”  t  contract  o»  chant  ituSSTSTii 

/3  /  MDA-9j33— 7  9~C-fi2p2 


»■  FtmrOHHlHG  0»GAN|2ATi0N  NMI  NMD  AODNISS 

INSTITUTE  FOR  DEFENSE  ANALYSES 
400  Army-Navy  Drive 
Arlington,  Virginia  22202 


i0  »«OG"Am  EuCnCnT  PROJECT  T  *S« 
ARCA  4  «OR<  UNIT  NUMICRS 


Task  T-136 


"1 coMTMpj,LiMO  orricc  m«»»*  >«o  <oo«tsi  .  -  V .  1  -«•  *t»onT  oats 

DUSD  (Research  and  Advanced  Technology)  j  J|  ^pMHMM(8M979 
The  Pentagon,  Washington,  D.C.  20301  J.  ■']  1^zsssSsr«rfteSi p— — »  « 

Z  28 

TV  MONI^OAimiS  iGlwCT  77577  aSoRIM/ITT/JmSm  In«  cZtfrollm#  Qitrr*)  ’5  SfCuAlTvCuM-ii.ro/ 

Defense  Advanced  Research  Projects  Agency  TT,TnT  . 

1400  Wilson  Boulevard  UNCLASSIFIED _ 

Arlington,  Virginia  222C9  •  :  ...  '»•  8?ftou”'*,CA«5«  00”,i"*° 


MiOliTRl»UTiON  ft  ATCmCN  T  r*#« 


1  S«.  OCCU  ASSiPfCATlON  CO«NGA*0(MG 

>CMCOuki  N/A _ 


Approved  for  public  release;  distribution  unlimited. 


[7r  OlSTNlluTlON  STATlMCHT  (oi  'ft*  iMirMf  w*r*r««l  1*  Blurb  20.  It  OUtmrmmt  Irom  Bmoortj 


ELECTE 
MAR  5  I960 


SU**Uf*CNTA*V  NOTES 


UfNlNT 


wa  '  11  ,M -d-£ 


If.  KCT  VOPOI  (Conn  otto  on  roooroo  siOo  it  notomoory  and  iBmnttty  by  blotb  nxmtbot) 

forward-looking  infrared  systems,  atmospheres,  transmittance,  aerosols, 
attenuation,  signal-to-noise  ratio,  detection,  recognition, 
range  (distance) 

^<1  ^e^muMR  an  rt»NM  «id>  ft  moeobbotr  onp  lOmnnly  by  biotb  numbor) 

Forward-looking  infrared  (FLIR)  sensor  performance  range  estimates 
over  a  horizontal  path  can  be  made  by  using  a  rapid  estimation  procedure. 
These  estimates  can  be  calculated  while  in  the  field  by  using  a  hand 
calculator  for  several  different  FLIR  systems,  atmospheric  transmission 
conditions,  and  levels  of  difficulty  of  the  visual  task. 

\ 


00  U73  EDITION  Of  '  NOW  *4  I*  OMOLtT* 


UNCLASSIFIED 


SCCUNITV  CLAJJIAiCAT'ON  or  >ll  »*<ll  r'.im  Fnl*:*! 


y-titU' 


IDA  PAPER  P-1419 


FIELD  MANUAL  TO  DETERMINE  DETECTION 
OR  RECOGNITION  RANGE  OF  A  FLIR  SENSOR 


Lynne  N.  Seekamp 


September  1979 


INSTITUTE  FOR  DEFENSE  ANALYSES 
SCIENCE  AND  TECHNOLOGY  DIVISION 
400  Army-Navy  Drive,  Arlington,  Virginia  22202 


Contract  MDA  903  79  C  0202 
Task  T-136 


ACKNOWLEDGMENTS 


The  methodology  for  the  rapid  calculation  procedure  .to 
predict  FLIR  performance  discussed  in  this  paper  was  developed 
by  Dr.  Robert  E.  Roberts  of  the  Institute  for  Defense  Analyses. 
The  author  thanks  R.E.  Roberts,  L.M.  Biberman  and  M.L.  Sullivan 
of  the  Institute  for  Defense  Analyses  for  their  useful  discus¬ 
sions  and  contributions  to  this  paper. 


RE:  Classified  references,  distribut¬ 
ion  unlimited- 

Oelete  classified  references  per  Mr. 
Walter  Hanley,  IDA/Security 


ii 


ACCESSION  for  I. 

NTI$ 

ooc 

Buff  Section  Ol 

UNANNOUNCED 

□ 

iUSTIFICATION 

nr 

|  WSTWBUTIOH/ArjBUWUn  OODCS  1 

I  Dlst.  AVAIL  and/M 

CONTENTS 


Acknowledgments 


ii 


I.  INTRODUCTION 


II .  METHODOLOGY 


III.  INPUT  DATA 

A.  FLIR  System  Specifications 

B.  Task  Level  Factors 

C.  Target  Specifications 

D.  Atmospheric  Factors 


11 

12 

13 

15 

16 


IV.  AN  EXAMPLE  OF  THE  CALCULATION  PROCEDURE 


References 


25 


iii 


I.  INTRODUCTION 


This  publication  is  designed  to  be  a  concise  manual  for 
assessing  forward-looking  infrared  (FLIR)  sensor  performance 
ranges  under  a  variety  of  conditions  and  thus  is  directed 
toward  the  operator  in  the  field.  The  purpose  of  this  publi¬ 
cation  is  to  present  a  summary  of  the  methodology  used  to  de¬ 
rive  the  FLIR  performance  rapid  estimation  procedure,  present 
tables  of  input  values  for  several  different  FLIR  systems, 
atmospheric  transmission  conditions  and  levels  of  difficulty 
of  the  visual  task,  and  show  an  example  of  the  calculation  pro¬ 
cedure.  In  this  paper  we  will  use  what  we  believe  are  the  best 
data  bases  and  models  currently  available.  However,  uncertain¬ 
ties  still  exist  regarding  some  of  these  data,  particularly 
with  respect  to  the  prediction  of  aerosol  extinction.  These 
uncertainties  will  be  discussed  in  the  appropriate  sections. 

A  previous  publication  described  a  computer  code  developed 
to  model  the  performance  of  a  FLIR  sensor  (Ref.  1).  The  com¬ 
puter  code  (Program  FLIR)  was  used  to  calculate  the  probabili¬ 
ties  of  detection  and  recognition  of  a  target  by  an  observer 
using  a  FLIR  sensor.  Since  that  time  a  rapid  approximation  for 
calculating  the  range  at  various  probabilities  of  detection  or 
recognition  for  a  given  FLIR  system  has  been  developed.  We 
felt  that  it  would  be  useful  to  publish  a  concise  manual  that 
could  be  used  in  the  field  as  a  guide  for  making  quick  estimates 
of  FLIR  performance.  It  should  be  noted  that  the  procedure  out¬ 
lined  here  is  applicable  only  to  estimating  FLIR  performance 
over  horizontal  paths.  The  equation  used  to  estimate  FLIR  per¬ 
formance  was  derived  by  R.E.  Roberts  of  the  Institute  for  De¬ 
fense  Analyses  (Refs.  2,  3). 


1 


The  calculation  method  presented  here  is  designed  to  be 
used  with  a  hand  calculator  when  a  fast  determination  of  range 
at  various  probabilities  of  detection  or  recognition  is  required. 


II.  METHODOLOGY 


An  equation  has  been  derived  which  directly  shows  the  range 
at  50  percent  probability  (as  well  as  at  other  probabilities)  as 
a  function  of  target,  system,  environment,  and  task  parameters. 
An  explanation  of  the  derivation  of  this  equation  and  the  vali¬ 
dation  are  given  in  IDA  Paper  P-1284,  A  Simplified  Approach  to 
Analyses  of  Infrared  Sensor  Performance  Versus  Weather:  Theory 
and  Application  to  the  Hannover  Data  Base  (U)  (Ref.  2).  A  brief 
discussion  of  the  methodology  will  be  reviewed  below. 

The  most  common  parameter  used  to  characterize  FLIR  per¬ 
formance  is  the  minimum  resolvable  temperature  (MRT).  Usually 
this  represents  a  plot  of  the  values  of  minimum  resolvable  tem¬ 
perature  difference  between  a  pattern  of  four  identical  bars  and 
the  three  spaces  between  them  for  each  of  a  number  of  spatial 
frequencies.  The  bars  represent  a  blackbody  source  of  tempera¬ 
ture  T  +  AT,  where  T  is  the  background  temperature  (in  degrees 
Kelvin)  and  AT  is  the  difference  between  the  background  tempera¬ 
ture  and  the  target  temperature.  The  spaces  represent  the  back¬ 
ground  temperature  T.  MRT  then  is  the  minimum  AT  that  a  stand¬ 
ard  observer  can  resolve  through  a  given  FLIR.  The  four  bars 
and  three  equal-sized  spaces  form  a  square,  and  thus  the  aspect 
(length: width)  ratio  of  each  bar  and  space  is  7:1  (Fig.  1). 

J.  Johnson  has  shown  that  many  objects  may  be  represented 
by  pairs  of  black  and  white  bars  inside  a  square  (Ref.  4).  He 
related  the  number  of  line  pairs  of  a  bar  chart,  where  the  pairs 
fit  inside  the  minor  dimension  of  an  object,  with  the  ability 
of  an  observer  to  detect  (resolve  one  pair  of  bars)  or  recog¬ 
nize  (resolve  four  pairs  of  bars)  the  object.  The  task  level 


TEMPERATURE  =  T 


TEMPERATURE  =  T  +  A  T 


FIGURE  1.  Standard  four-bar  MRT  test  pattern. 

factor  y  is  the  number  of  line  pairs;  for  example,  for  detec¬ 
tion  y  =  1  and  for  recognition  y  *  Johnson's  conclusion 

was  based  on  the  assumption  that  the  object  or  target  and  the 
equivalent  bar  pattern  were  of  the  same  size  and  contrast  and 
at  the  same  distance. 

F.A.  Rosell  introduced  the  concept  of  aspect  corrections 
for  targets  whose  shapes  are  significantly  different  from  the 
square  used  in  the  standard  equivalent  bar  patterns  (Ref.  5). 

As  a  result,  the  laboratory-measured  value  of  MRT  using  the 
standard  four-bar  square  test  pattern  with  an  aspect  ratio  of 
7:1  is  divided  by  /e/7  in  order  to  obtain  an  aspect-corrected 
value  of  MRT.  The  letter  e  is  the  length-to-width  ratio  of  a 
single  resolution  bar  in  the  pattern  which  represents  the  actual 
target  rectangular  outline.  Figure  2  shows  an  automobile  with 
major  dimension  (length)  approximately  twice  the  minor  dimen¬ 
sion  (height)  and  the  corresponding  bar  patterns  for  detection 
and  recognition.  In  this  figure  the  aspect  ratio  of  a  single 
bar  for  detection  is  4:1  and  that  for  recognition  is  16:1. 


4 


DETECTION 


FIGURE  2.  Object  image  and  corresponding  bar  patterns 
for  detection  and  recognition. 


The  temperature  difference  between  a  target  and  its  back¬ 
ground  is  represented  by  AT  (in  degrees  Kelvin).  However,  at  a 
distance  the  apparent  temperature  differential  between  the  tar¬ 
get  and  the  background  will  be  less  because  of  atmospheric 
attenuation  of  the  target  radiation.  The  atmospherically  de¬ 
graded  thermal  contrast  is  obtained  by  multiplying  AT  by  the 
atmospheric  transmission  fatm- 

The  ratio  of  the  atmospherically  degraded  thermal  contrast 
to  the  aspect-corrected  value  of  MRT  is  a  normalized  signal-to- 
noise  ratio  (SNR).  Thus,  the  normalized  SNR  is  given  by  the 
expression : 


AT  t 


snrn  = 


atm 


MRT//e77 


(1) 


Equation  1  is  the  fundamental  expression  used  in  PLIR  perform¬ 
ance  modeling. 

It  has  been  shown  that  the  probability  of  carrying  out  a 
given  task  of  detection  or  recognition  is  related  to  the  nor¬ 
malized  SNR  as  indicated  In  Fig.  3-  These  data  were  generated 


5 


0.25  0.50  0.75  1.00  1.25  1.50  1.75  2.00 


SNRN  = 


mrt/v^T? 


FIGURE  3.  Probability  of  detection  or  recognition  versus 
normalized  s i g na 1  -  to- no i se  ratio. 


theoretically  and  from  a  series  of  experiments  using  many  ob¬ 
servers  (Ref.  6).  The  SNR  is  normalized  in  such  a  way  that 
when  SNR,,  =  1.0  the  probability  of  achieving  the  task  (detec¬ 
tion  or  recognition)  is  50  percent.  Other  values  of  SNRTJ  will 
yield  different  probabilities  of  achieving  the  task  (Fig.  3). 

The  solution  to  Sc.  1  for  performance  range  is  greatly 
simplified  by  making  the  following  assumptions. 

First,  atmospheric  transmission  often  may  be  approximated 
by  Beer's  law  for  narrow  spectral  intervals  or  spectral  regions 
where  there  is  a  relatively  weak  range  dependence  such  as  for 
the  water  vapor  continuum  or  aerosol  extinction.  If  we  assume 
a  3eer's  law  dependence,  we  obtain 


-6  ,  R 

t  =  e  atm 
atm 


6 


where  i\ltm  la  the  extinction  coefficient:  of  the  atmosphere  and 

It  la  range  (usually  in  kilometers).  The  coefficient  g  ,  Is 
°  at  in 

the  sum  of  the  extinction* coefficients  due  to  molecular  absorp¬ 
tion,  water  vapor  continuum,  and  aerosols: 

g  ®  g  ,  +  g , ,  .  ,  +  g 

atm  mol  H-.0  cent  aer 

Look-up  tables  of  ^  +  3n-,o  coiit  aro  pf'-'acnt. ed  later.  It.  is 
evident  from  these  tables  of  extinct  ion  coefficients  for  differ¬ 
ent  path  lengths  that  g^  ^  has  a  weak  dependence  on  range.  The 
example  provided  will  show  how  to  deal  with  this  problem  in  a 
practical  way.  The  dominant  causes  of  atmospheric  attenuation, 
namely  aerosols  and  the  water  vapor  continuum,  are  relatively 
Independent  of  range. 

We  next  assume  that  the  system  MKT  also  can  be  adequately 
approximated  by  an  exponential  function: 

v 

MKT  =  MKT  e  -•v  .  (j) 

o 

Figure  shows  MKT  ,  g  ,  and  v  on  a  plot  of  the  logarithm  of 

O  io  y  o 

minimum  resolvable  temperature  (MKT)  versus  spatial  frequency 
In  eye  les/tni  1  1  I  rad  Ian,  where  MKT  (  Is  the  y-Intercept  of  the  re¬ 
gression  line,  g.  ,  Is  the  system  ext  lnct  Ion  coefficient  ,  and 

i>y  o 

v  is  the  spatial  I’requency  In  eye  les/ml  1  1  1  rad  I  an  .  If  the  s  1  s.e 
(:’-)  of  the  target  (the  minor  dimension  In  meters  1  and  the  task 
1  ovel  factor  t>)  defined  according,  to  the  Johnson  equlvalent- 
bar-p.at  tern  criteria  are  known,  then  we*  have 


v  - 


a  sea  Ling. 


Th  1  s.  equal  Ion  pr  'V  Ides. 


fad  or  for  l  tie  abs.c  I  ssa  I  n 


MRT 


V  =  FREQUENCY  (cycles/mrad) 


FIGURE  4.  Sample  MRT. 


8 


Substituting  ^  R  for  v  in 


MRT  =  MRT 

o 


Eq . 

I 

S 


3  yields 


6 


sys 


(4) 


The  performance  range  in  kilometers  at  the  50  percent  confidence 
level  is  then  obtained  by  substituting  Eqs.  2  and  4  into  Eq.  1 
for  eafcm  and  MRT,  respectively,  and  by  setting  SNR^  =  1,  which 
by  definition  gives  a  probability  of  50  percent  of  achieving 
the  task  (Fig.  3 ) : 


1  = 


AT  e 


_SatmR 


MRT  e 
o 


X  8  R 

S  sys  //£77 


Equation  5  is  then  solved  for  R: 

AT  /S7T 


In 


MRT. 


R  = 


B  ^  +  £  B 

atm  S  sys 


(5) 


(6) 


Table  1  shows  values  of  MRT  and  B_,,„  For  two  FLIR  systems 

o  sys 

representing  1974  and  1978-79  technology  devices.  The  coeffi- 
cient  of  determination  r“  is  based  on  a  regression  on  Eq.  3  and 
is  an  indication  that  the  two  parameters  MRT  and  B„,re  are  valid 

o  syb 

descriptors  of  MRT. 

TABLE  1.  REPRESENTATIVE  MRT  AND  B.v<;  VALUES 


8-12  um  FLIR 


1974  1978-79 

_ T  echnol  ogy _ Technology 

MRTq  (°K)  0.0254  0.0112 

B  (mrad/cycle)  0.996  0.633 

sys 

r2  0.99  0.99 


3-5  um  FLIR 
1974 

T echnol ogy 
0.0171 
1  .006 
0.99 


Q 


Equation  6  is  the  range  performance  at  50  percent  confidence 
level  (SNR.,  =1).  To  determine  range  performance  at  the  10  per- 
cent  confidence  level  and  the  90  percent  confidence  level  refer 
to  Fig.  3.  The  value  on  the  normalized  curve  for  a  probability 
of  10  percent  is  SNR^  =  0.5,  and  for  a  probability  of  90  percent 
it  is  SNR^  =  1.5*  Therefore,  rewriting  Eq.  6,  we  get 


R 


P 


AT  /uj\ 

mrtq  ) 


sys 


(7) 


where  Rp  is  the  expected  range  for  probability  p  of  detection 
or  recognition  and  kp  is  the  normalized  SNR  for  probability  p. 
When,  for  example, 


p  =  0.90,  kp  =  1.5; 

p  =  0. 50,  kp  =  1. 0; 

p  =  0.10,  kp  =  0.5. 

Validation  of  Eq.  7  has  been  shown  in  Ref.  2. 

The  next  section  presents  some  data  that  can  be  used  as 
input  to  Eq.  7. 


1C 


III.  INPUT  DATA 


This  section  contains  sample  values  for  the  parameters  in 
Eq.  7-  These  data  are  inputs  that  we  have  used  at  IDA  and  rep¬ 
resent  our  best  understanding  of  the  problem.  These  data  are 
only  intended  to  be  examples  of  the  inputs  necessary  to  calcu¬ 
late  the  range  for  certain  probabilities  of  detection  and  recog¬ 
nition  using  the  method  discussed  in  this  paper. 

The  sample  input  values  are  arranged  by  type  of  parameter: 
FLIR  system  specifications ,  task  difficulty  factors,  target 
specifications,  and  atmospheric  or  environmental  factors.  A 
discussion  is  presented  wherever  it  is  necessary  to  comment  on 
the  source  of  the  data  and  on  reservations  we  may  have  about 
the  accuracy  or  validity  of  the  models  used. 


11 


A.  FLIR  SYSTEM  SPECIFICATIONS 


Parameter  Description 

MRT  y-intercept  of  the  regression  line 

'  fitted  to  the  values  of  the  MRT  in  the 

specifications  of  the  particular  FLIR 
sensor  of  Interest  (Fig.  4).  This 
relates  to  the  overall  FLIR  sensitiv¬ 
ity. 


S  Slope  of  the  regression  line  fitted  to 

y  the  values  of  the  MRT  in  the  specifica¬ 

tions  of  the  particular  FLIR  sensor  of 
interest  (Fig.  4).  This  relates  to 
the  FLIR  resolution. 


Sample  Values 

8-12  ym  FLIR 
1974 _ 1978-79 

MRTq  (°K)  0.0254  0.0112 

6syg  (mrad/ cycle )  0.996  0.633 


3-5  um  FLIR 
1974 

0.0171 

1.006 


B.  TASK  LEVEL  FACTORS 


Parameter 


e 


Y 


k 


P 


Description 

Bar  aspect  ratio  (length-to-width  ratio  of 
a  single  bar)  of  line  pairs  placed  across 
a  rectangular  cross  section  of  the  target 
where  width  of  the  rectangle  is  equal  to 
the  minor  dimension  of  the  target  and 
length  is  the  major  dimension  (Fig.  5). 

Task  level  factor,  the  number  of  bar 
chart  line  pairs  per  minor  dimension 
of  the  object. 

Scaling  factor  for  determining  range  for 
probability  of  detection  or  recognition 
equal  to  p. 


LENGTH 


MINOR 

DIMENSION 

OF 

TAR6ET 


4M-7H 


DETECTION 


RECOGNITION 


FIGURE  5.  Bar  test  pattern  for  the  front  aspect  of  a  tank. 


Pi s cuss  1  on 

Figure  5  shows  the  bar  test  pattern  used  for  the  front 
aspect  of  a  tank.  For  the  front  aspect  of  a  tank  the  rectangle 
is  assumed  to  be  approximately  square  so  that  the  length  and 


13 


width  are  both  equal  to  the  minor  dimension  of  the  tank  (the 
height).  However,  for  the  side  aspect  of  a  tank  the  length 
would  be  increased,  while  the  width  of  the  rectangle  would  re¬ 
main  equal  to  the  height  of  the  tank.  For  our  sample  values 
we  assume  a  tank  that  has  length  equal  to  approximately  three 
times  its  height. 

Sample  Values 

The  number  of  bars  placed  across  the  rectangle  has  been 
determined  to  be  2  for  detection  and  8  for  recognition  (Ref.  4), 
as  shown  in  Fig.  5-  Examples  of  aspect  ratios  are  the  following 


Front  Aspect  Tank 
Width=3m  Heiqht=3m 

Side  Aspect  Tank 
Lenqth=9m  Heiqht=3m 

Rectangle  Aspect 
Ratio 

1 

:  1 

3: 

1 

Detecti on 

Re  coqni ti on 

Detecti on 

Recoqni ti on 

y  (Number  of 

Line  Pairs) 

1 

4 

1 

4 

Number  of  Bars 

2 

8 

2 

8 

Bar  Aspect 

Ratio 

2:1 

8:  1 

6:1 

24:1 

e 

2 

8 

6 

24 

Normalized  SNR 
for  Probability 
p  ( k0)  from 

Fig.  3: 

k .  90 

1.5 

1.5 

1.5 

1.5 

k.50 

1.0 

1.0 

1.0 

1.0 

k.  10 

0.5 

0.5 

0.5 

0.5 

14 


C.  TARGET  SPECIFICATIONS 


DETERMINE  TARGET  MINOR 
DIMENSION  AND  ATQ 


4-3-791 

Parameter  Description 

S  Target  size  is  the  minor  dimension  of 

the  target  expressed  in  meters. 

AT  Effective  temperature  difference  (thermal 

contrast)  between  the  target  and  the  back¬ 
ground  in  degrees  Kelvin. 

Sample  Values  for  a  Tank  Target 

S  The  minor  dimension  of  a  tank  is  its 

height.  Generally  we  have  been  using 
3  meters  as  the  height  of  U.S.  tanks; 
therefore  S  =  3  meters. 

Front  Aspect  Tank  Side  Aspect  Tank 

AT  £2°K  j»6°K 


15 


D.  ATMOSPHERIC  FACTORS 


Parameter  Description 

8  .  Extinction  coefficient  of  the  atmosphere 

(km-1),  which  is  the  sum  of  the  extinctions 
due  to  molecular  absorption,  water  vapor 
continuum,  and  aerosols: 

8  ,  =  8  ,  +  8n  .  ,  +  8 

atm  mol  H2O  cont  aer 

Discussion 

This  atmospheric  parameter  is  complex  and  requires  a  brief 
review  in  this  section.  A  detailed  discussion  of  the  knowledge 
to  date  on  this  subject  can  be  found  in  Ref.  7,  Chapter  III, 
"Atmospheric  Effects  on  Infrared  Systems,"  by  J.B.  Goodell  and 
R.E.  Roberts. 


As  indicated  above,  Batm  comprises  three  components: 


ex¬ 


tinction  due  to  molecular  absorption,  water  vapor  continuum, 
and  aerosols.  For  this  discussion  of  extinction  Beer's  law  is 


assumed  (t  , 
atm 

atmosphere  (t 


=  .-BR 


=  e 


),  and  therefore  transmission  through  the 


atm 


)  will  be  calculated.  Once 


t  ,  is  determined 
atm 


and  range  R  is  selected,  8  can  be  obtained. 


The  two  components  molecular  absorption  and  water  vapor 
continuum  are  computed  in  computer  code  LOWTRAN  3b,  developed 
by  the  Air  Force  Geophysics  Laboratory  to  calculate  atmospheric 
transmittances  (Ref.  8).  This  computer  code  is  widely  accepted 
as  the  best  model  for  computing  molecular  band  absorption. 
LOWTRAN  3b  also  calculates  the  aerosol  component,  but  this  will 
be  discussed  separately. 


The  following  tables  enable  the  user  to  obtain  a  value  for 
the  molecular  extinction  coefficient  8,  given  the  atmospheric 
temperature  T&  (°C),  the  dew  point  T^  (°C),  and  the  range  R 
(km).  The  values  of  8  are  derived  from  the  molecular  and  water 
vapor  continuum  components  of  atmospheric  transmission  as  com¬ 
puted  using  the  LOWTRAN  3b  code  for  a  horizontal  sea-level  path, 
and  a  weighting  function  corresponding  to  a  blackbody  source  at 
10°C. 


16 


p 


Table  2  presents  extinctions  for  the  3-5  ym  band.  Note 
that  atmospheric  temperature  Ta  is  not  needed  in  order  to  find 
&3_q;  only  the  dew  point  Tdp  and  range  R  are  necessary.  For  a 
given  Tdo>  a  change  in  T&  does  not  produce  any  appreciable  dif¬ 
ference  in  63  3 .  Therefore,  all  ranges  appear  together  in  one 
table. 

For  the  8-12  um  band,  both  T  and  T,  are  needed  to  find 

a  dp 

the  appropriate  Sg_12>  so  a  separate  table  is  provided  for  each 
range.  Tables  3  through  6  correspond  to  ranges  of  2,  4,  8,  and 
16  km,  respectively. 

The  relationship  between  t  tm  and  R  assuming  Beer's  law 
(Eq.  2)  for  the  8-12  ym  band  is  not  as  strong  as  for  the  3-5  ym 
band.  Therefore,  the  range  R  to  be  used  in  looking  up  the 
appropriate  8mol  +  Bh20  cont  in  Tables  3  through  6  should  be 
estimated.  Using  the  estimated  value  of  R,  the  value  for  ex¬ 
tinction  is  selected,  and  Eq .  7  is  solved  for  R.  If  R  is  some¬ 
what  different  from  the  R  used  to  look  up  the  extinction  value, 
Eq.  7  should  be  solved  again  using  the  more  appropriate  value 
for  extinction.  This  process  is  shown  in  the  example  presented 
at  the  end  of  the  paper. 

For  values  of  Ta,  Tdp,  or  R  other  than  those  given,  a 
linear  interpolation  scheme  between  the  nearest  given  values 
will  yield  a  fairly  accurate  result. 

Again,  these  values  of  6  account  for  only  the  molecular 
and  water  vapor  continuum  components  of  extinction.  To  obtain 
a  realistic  estimate  of  detector  performance,  the  aerosol  com¬ 
ponent  of  extinction  must  be  taken  into  account  as  well. 

There  are  many  uncertainties  regarding  aerosol  effects 
on  FLIR  systems.  LOWTRAN  3b  contains  aerosol  models  whose 
validity  is  in  doubt.  However,  the  data  given  in  LOWTRAN  3b 
are  used  for  our  purposes.  We  have  derived  approximations  based 
on  the  LOWTRAN  3b  aerosol  data.  The  derivation  of  these  aerosol 


17 


TABLE  2.  3-5  ym  MOLECULAR  PLUS  H2O  CONTINUUM  EXTINCTION 

COEFFICIENT  FOR  TARGET  TEMPERATURE  OF  10°C 


'dP 

7*rr 

mm 

-20 

nu 

BQ 

-5 

<5 

5 

msm 

■H 

20 

m 

30 

35 

4(3 

0.5 

.602 

.643 

.691 

.  745 

.810 

.880 

.956 

1.038 

1.128 

1.225 

1  .  331 

1.439 

1.557 

1.0 

.381 

.41 1 

.446 

.486 

.529 

.576 

.629 

.  685 

.  747 

.810 

.875 

.947 

1.024 

.246 

.268 

.292 

.  318 

.  349 

.381 

.415 

.452 

.490 

.532 

.576 

.622 

.672 

.161 

.176 

.192 

.211 

.231 

.252 

.275 

.299 

.  325 

.  351 

.379 

.410 

.443 

8.0 

.107 

.117 

.129 

.141 

.154 

.168 

.183 

.199 

.215 

.233 

.252 

.275 

.  300 

16.0 

.072 

.079 

.087 

.095 

.104 

.113 

.123 

.133 

.145 

.158 

.173 

.190 

.210 

32.0 

.049 

.054 

.059 

.065 

.070 

.084 

.091 

.100 

.111 

.122 

.136 

.151 

TABLE  3.  8-12  ym  MOLECULAR  PLUS  H2O  CONTINUUM  EXTINCTION 

COEFFICIENT  FOR  TARGET  TEMPERATURE  OF  10°C  AND 
PATH  LENGTH  R  =  2.0  km 


.  .  . _ LtJ 

7T°cl 

Ta  (°C) 

-20 

wsm 

-10 

-5 

0 

5 

10 

15 

20 

mm 

30 

35 

40 

-20 

.039 

ap:- 

.039 

.047 

.038 

.047 

.058 

.038 

.046 

.057 

.075 

.038 

.045 

.056 

.073 

.099 

5 

.037 

.044 

.055 

.071 

.096 

.  1  37 

.037 

.044 

.054 

.069 

.093 

.131 

.195 

15 

.037 

.043 

.053 

.067 

.090 

.126 

.185 

.283 

20 

.036 

.043 

.052 

.066 

.087 

.121 

.176 

.267 

.  41  7 

25 

.036 

.042 

.052 

.064 

.085 

.117 

.168 

.254 

.  393 

.616 

30 

.036 

.042 

.051 

.063 

.082 

.113 

.161 

.241 

.  371 

.  579 

.907 

35 

.035 

.042 

.050 

.062 

.081 

.230 

.  352 

.547 

.  852 

1  .  323 

40 

.035 

.041 

.050 

.062 

.079 

.107 

.150 

.  334 

.516 

.805 

1  .  244 

1  .  908 

TABLE  4.  8-12  pm  MOLECULAR  PLUS  H2O  CONTINUUM  EXTINCTION 

COEFFICIENT  FOR  TARGET  TEMPERATURE  OF  10°C  AND 
PATH  LENGTH  R=4.0  km 


TABLE  5.  8-12  pm  MOLECULAR  PLUS  H2O  CONTINUUM  EXTINCTION 

COEFFICIENT  FOR  TARGET  TEMPERATURE  OF  10°C  AND 
PATH  LENGTH  R=8.0  km 


A 


1? 


TABLE  6.  8-12  um  MOLECULAR  PLUS  HoO  CONTINUUM  EXTINCTION 

COEFFICIENT  FOR  TARGET  TEMPERATURE  OF  10°C  AND 
PATH  LENGTH  R=16.0  km 


model  approximations  is  discussed  at  length  in  Ref.  7.  In  addi¬ 
tion  to  the  LOWTRAN  3b  aerosol  models  (maritime,  urban  and  rural) 
an  aerosol  model  for  dry  climates  has  been  developed  here.  Vis¬ 
ibility  (VIS)  in  kilometers  is  the  only  input  value  needed  to 
determine  S  by  using  the  approximations  (Table  7). 

cLGiv 

The  uncertainties  regarding  the  aerosol  models  must  be 
emphasized.  There  are  three  basic  caveats  to  keep  in  mind: 

1.  The  visibility  required  as  input  to  the  aerosol 
models  is  not  a  reliable  measure.  It  is  very 
common  for  two  people  to  get  different  values 
when  measuring  visibility.  For  a  more  detailed 
discussion  see  Ref.  7. 

2.  It  is  unrealistic  to  expect  any  simple  scaling 
model  will  pertain  to  all  atmospheric  conditions. 

For  example,  the  continental  model  may  not  be 
appropriate  for  all  continental  atmospheric  con¬ 
ditions;  for  some  limited-visibility  conditions 


2C 


over  the  continent,  the  maritime  model  la  a  better 
approximation  of  those  conditions. 


3.  The  models  in  Table  7  are  for  ground-level  paths. 
For  air-to-ground  cases  significant  differences 
occur  due  to  vertical  structure  which  cannot  he 
predicted  by  these  models. 


TABLE  7. 

EQUATIONS  TO  APPROXIMATE 

^aer 

Aerosol  Model 

8-12  um 

3-5  um 

Mari  time 

0.85 

2.24 

VTT~ 

FIT" 

Rura  1 

0.43 

0.42 

FTT- 

wr 

Urban 

0.41 

0.60 

FIT" 

FIT- 

Dry 

0.95 

1  .  76 

FIT" 

FIT" 

Thus,  |3  tm  may  be  determined  by  using  the  appropriate  valu 
obtained  from  Tables  2-b  (B[uo^  +  fn^O  cont  ^  at'd  adding  it.  to 
,  calculated  by  using  the  approximations  for  different 

el  t?  I 

atmospheric  conditions  given  in  Table 


IV.  AN  EXAMPLE  OF  THE  CALCULATION  PROCEDURE 


I.  Assume  the  following  problem  specifications: 

A.  FLIR  System  Specifications  for  a  1974  Generation  8-12  ym 
FLIR: 

MRTQ  =  0.025^  °K 

8  =  O.996  mrad/cycle 

sy  s 

B.  Task  Level  Factors  for  a  Front  Aspect  Tank  3  Meters  in 
Height  for  Probability  of  Detection  of  50  Percent: 

Y  =  1 

e  =  2 

kp  =  k. 50  =  1,0 


C.  Target  Specifications: 

S  =  3  meters 

AT  =  2°K 


D. 


Atmospheric  Factors  when  Visibility  =  2  km  under  Mari¬ 
time  Environment  with  T&  =  10°C  and  T^  =  0 °C 


0.85  _  0.85  _ 


2.0  km 


=  0.425  km 


2.  Estimate  range  (R  )  to  use  in  looking  up  appropri- 
ate  Smol  +  V  cont  in  Tables  2-6: 


22 


Rest 


AT  /F7T\ 
MRTq  J 


S  +  X  B 
aer  S  sys 


0.425  +  j  (0.996) 


_  3.740 
0.757 


=  4 . 9 


Thus  we  will  use  Table  4  to  look  up  6  n  ^ 

moj.  H2O  cont 

since  it  is  for  range  =  4  km,  which  is  the  closest  range  to 
4.9  km  in  the  tables  presented  in  this  report.  For  T.  =  10°C 

cL 

and  Tdp  =  0  C  and  R  =  4  km,  we  get  Bmol  +  BHoQ  cont  =  km 

Thus,  Batm  *  0.425  +  0.078  =  0.503  km-1. 

(Note:  A  linear  interpolation  could  be  done  between  values 
found  in  Table  4,  where  range  =  4  km,  and  Table  5,  where  range 
=  8  km,  to  get  Bmol  +  B^  CQnfc  =  0.076  km"1.) 


II.  Solve  Eq.  7  for  all  the  input  values  given: 


R 


.50 


In 

0.503 


/l  2  JT7 t\ 

\1  0.025V 

+  j  (0.996) 


„  3-740 
0.335 

=  4.48 


23 


Thus,  the  range  at  50  percent  probability  of  detection 
for  the  given  PLIR  and  under  the  given  atmospheric  conditions 
is  4.5  km. 


REFERENCES 


1.  Institute  for  Defense  Analyses,  Effect  of  Weather  at 

Hannover ,  Federal  Republic  of  Germany ,  on  Performance  of 
Electrooptical  Imaging  Systems:  The  Calculation  Method¬ 
ology  for  a  FLIR  Using  a  FORTRAN  Program ,  IDA  Note  N-842, 
L.N.  Seekamp,  August  1977. 


3. 


4.  J.  Johnson,  paper  presented  at  Image  Intensifier  Symposium, 
Ft.  Belvoir,  Virginia,  6-9  October  19-58. 

5.  F.A.  Rosell  and  R.H.  Willson,  "Recent  Psychophysical  Experi¬ 
ments  and  the  Display  Signal-to-Noise  Ratio  Concept,"  Chap¬ 
ter  5  in  L.M.  Biberman,  ed..  Perception  of  Displayed  Infor¬ 
mation,  Plenum  Press,  New  York,  1973,  pp.  167-232. 

6.  R.L.  Legault,  "Visual  Detection  Process  for  Electrooptical 
Images:  Man — The  Final  Stage  of  an  Electrooptical  Imaging 
System,"  Chapter  4  in  L.M.  Biberman  and  S.  Nudelman,  eds., 
Photoelectronic  Imaging  Devices,  Vol.  1,  Physical  Processes 
and  Methods  of  Analysis ,  Plenum  Press,  New  York,  1971, 

pp.  69-86. 

7.  U.S.  Navy  Electrooptical  Technology  Program  Office/  Naval 
Research  Laboratory,  The  Fundamentals  of  Thermal  Imaging 
Systems,  EOTPO  Report  46/NRL  Report  8311,  F.A.  Rosell  and 
G.  Harvey,  eds..  May  10,  1979- 

3.  Air  Force  Geophysics  Laboratory,  Atmospheric  Transmittance 
from  0.25  to  28.5  \xm :  Supplement  LOWTRAN  3b  (1976),  J.E.A. 
Selby,  E.P.  Shettle,  and  R.A.  McClatchey,  November  1976. 


25