AD-A 46 71 SUPEPORT FOR THE NAVAL RESEARCH LABRATORY PASSIV /E YE OM D MS NCS0V G CROIAVE CLUTTER ANALY.U) COMPUTER SCIENCES CORP d ~Aq;FE OT8 SC/TR84/6005 N000483C2316 / 179 N I flflEE........ K.
AD-A 46 71 SUPEPORT FOR THE NAVAL RESEARCH LABRATORY PASSIV /E
YE OM D MS NCS0V G CROIAVE CLUTTER ANALY.U) COMPUTER SCIENCES CORP
d ~Aq;FE OT8 SC/TR84/6005 N000483C2316 / 179 N
I flflEE........
K.
lii &- &128A 6
1111.2 11 68.II1.25 ___ W 11
MICROCOPY RESOLUTION TEST CHARTNATIONAL BUREAU OF STAN04DS1963-A
"Ib
CSCTR4I6 t
I JSUPPORT FOR THE NAVAL RESEARCHLABORATORY PASSIVE MICROWAVE CLUTTER
! ANALYSIS PROGRAM
FINAL CONTRACTOR REPORT
Prepared forDEPARTMENT OF THE NAVY
I0 NAVAL RESEARCH LABORATORYWashington, D.C.
CONTRACT N00014-83-C-2316Task Assignment 5104
L.OCTOBER 1964 andpl5c~It
Iis
COMPUTER SCIENCES CORPORATION L
COMPUTER SCIENCES CORPORATIONSYSTEM SCIENCES DIVISION (301) 58 9-154 5
,,, 8728 COLESVILLE ROAD • SILVER SPRING. MARYLAND 20910
October 11, 1984
Defense Technical Information CenterAttention: DDACameron Station, Building 5Alexandria, Virginia 22304-6145(703) 274-7633
Dear Sirs:
Enclosed are 12 copies of the unclassified, unlimited distributionFinal Contractor Report required by NRL Contract N00014-83-C-2316.
Yours truly,
R. A. Nieman
Project Manager
RAN:gls
Enclosures (12)
OFFICES IN PRINCIPAL CITIES THROUGHOUT THE WORLD
I
SUPPORT FOR THE NAVAL RESEARCH LABORATORYPASSIVE MICROWAVE CLUTTER ANALYSIS
PROGRAM - FINAL CONTRACTOR REPORT
Prepared for
THE NAVAL RESEARCH LABORATORY
By
COMPUTER SCIENCES CORPORATION
Under
Contract N00014-83-C-2316Task Assignment 5104 -
OCT 2 3 14
APrepared by: Approved by:
TV _ ____ -_ j X-T. Griffin 11' Date Dr. M. Allen Date
Reviewed by:
Dr. R. Nfeman Date
This dpc~fleft has be"~ approvedfor public releos. and uai%; itsdis.ibubto is unlirits.&
I I -
I
ABSTRACT
This document suimmarizes the data acquisition, reduction, and processingsupport CSC provided Code 7911 of the Naval Research Laboratory (NRL)under contract N00014-83-C-2316. The support for the NRLPassive Microwave Clutter Program included development of algorithmsfor processing and analyzing data obtained by a ground-based microwaveradiometer observing thermal radiation from various targets. Datareduction and analysis techniques are discussed.
K I
It
II
I as.
Ir
.O~OI 9
TABLE OF CONTENTS
Section 1 Introduction.......................................... 1-1
Section 2 - Microwave Temporal Clutter Dat~a Reduction Programs.... 2-1
2. RAA DaaRdcinPorm................. -2.2 RADDAT Data Reduction Program .............................. 2-1
2.3 RDDAT Data Reduction Program ............................... 2-2 rSection 3 - Microwave Temporal Clutter Analysis Programs ...........3-1
3.1 KADOUT Data Analysis Program ............................... 3-1
3.1.1 The Fast Fourier Transform Analysis Module ............3-3I3.1.2 The Temporal Structure Function Analysis Module .......3-43.1.3 Temperature Averaging and Differencing Analysis Modul..3-73.1.4 Auto-"correlation Analysis Module ......................3-7
I3.2 RDOUT Data Analysis ........................................ 3-10
13.3 Averaging and Normalization Programs ........................ 3-10
Appendix A - Software Documentation .............................. A-1
References.....................................................R-1
II
~LIST OF ILLUSTRATIONS
I FIGURE
3-1 Normalized Harm-Windowed Averaged Grass Scene FFT Spectrumfor 37 GHz Horizontal (H) Polarization ...................... 3-7
3-2 Background Scene Temporal Structure Versus Time ............. 3-6
3-3 Auto-correlation Graph: Auto-correlation of a KnownSignal Versus Record Number for 37 GHz (Horizontal andVertical Polarization) ..................................3-. 3-9
3-4 Normalized Hann-Windowed Enclosure FFT Spectrum:37 GHz H-Polarization Enclosure Signature .................. 3-13
3-5 Target Differenced Hann-Windowed FFT Spectrum for 37 GHzH-Polarization: This Figure shows the Grass BackgroundContribution to the FFT Spectrum .......................... 3-14
3-6 Normalized Enclosure Temporal Structure Versus Time ....... 3-15
3-7 Target Differenced Clutter Temporal Structure VersusTime .................................................... 3-16
3-8 Normalized Difference Temporal Structure Versus Time ..... 3-17
IIIIII I
! 4"I
I
t LIST OF TABLES
TABLE
1-1 Summary of Available Clutter Analysis Data ................. 1-2
1-2 FORTRAN Software For NRL Passive Microwave TemperalClutter Project ............................................ 1-5
3-1 Brightness Temperature Data Base for MicrowaveTemporal Clutter Analysis .................................. 3-2
3-2 Brightness Temperature Analysis for Sample Averagingand Temperature Differencing ............................... 3-8
3-3 Brightness Temperature Data Sets Used for TargetAveraging .................................................. 3-12
Iv
I
I
I
I! ~
I
SECTION 1 - INTRODUCTION
The Computer Sciences Corporation (CSC) supported the Passive MicrowaveClutter Analysis Program for Code 7911 (recently reorganized as Code
7780) of the Naval Research Laboratory (NRL) under contract N00014-83-C-2316from October 1, 1983, to August 31, 1984. This report summarizes thesupport provided by CSC on this contract.
The objective of the project is to evaluate the amount of clutter re-
jection required in order to reach instrumental noise limits for targetdetection in the presense of atmospheric and background clutter witha multiple beam step-stare passive microwave space surveillance system.Data to be analyzed were collected in October and November 1983 andFebruary 1984, using existing 37 gigahertz (GHz) radiometer and datacollection equipment. This hardware required little or no modification;it has been described in Reference 1. The clutter data analyzed are
comprised of long stares (approximately five minutes in duration)
by the radiometer at stationary targets and at the sky. Data obtainedduring the 1983 observing period were of targets located at NRL. Thetargets included asphalt, grass, water, metal foil, mirror, and the skyat zenith and the horizon. Data obtained during the 1984 observingperiod consisted of sky and absorber (enclosure) observations at thePatuxent River Naval Air Test Center (NATC-PAX) *near Lexington Park,Maryland. A summary of all observation and calibration data availablefor the Clutter Analysis project is presented in Table 1-1.
Computer programs for data reduction and analysis were written by CSCpersonnel to operate on the DECsystem-lO computer at NRL. A summaryand brief description of these programs is presented in Table 1-2.The reduction techniques are discussed in Section 2.
The target data have been statistically analyzed to obtain an estimateof temporal instrumentation and background noise levels. Four methodsof statistical analysis were utilized: Fourier transformation analysis,auto-correlation analysis, temporal structure analysis, and temperatureaveraging and differencing. A discussion of data analysis techniquesis presented in Section 3.
Finally, the computer programs developed for this project have beendocumented using standardized procedures. The in-code software docu-mentation is described in Appendix A.
1-1
.o V. 041 a b
CO 0 A i.0 .0 9
0~~ ~ ~ waob a4 .4V 1 4la$ 4 0 CL 41( 41 41:
A4 w. a 0 04o a 0r0 00-J hi 0hAi
cc ~4 $4 to v-4 to01'.co0 i4) 4.8 41 0 a 1 0 0
ri cch A1 co cc w .~
cct o ) V4$ a 41 M V. 41-a h4 "4 Mw a 0 r4 41 o "4 w w "q
hi 0 4.D 0 V- 1cc4 41 w 000 w .0 a) toc Ic0 w -H0 . o
.0. en 0.1W444 c 4 4 u CI 4
.241 0 0) aE-0 t :1 4 0a 0- 0 h *41
41 0 4 4 4 4E-i 00 a) 6) aa
I- -0 4 "4 r-4 " 4 M rpa~ 10 10 a U w U 0
co 00 Co P0 M 0. 0 "
0 04 UU a c 0 4oCD c W1 4
V4. 0 4 0 hi 0 0 - 0 a 0i hi 0414 w4 hi kJ $4 w k 4 w U4 ; 4 $
-r4 "4- 4,g " 4 -4 -H -M -H M - 4P6 -4 L2 is *a aa a a
5-. 00 41.0 en V.0 C4 "4 0 r- V- *
0 o GoC Go Go ~ 41 a a h
hi as
a ~9 4 00e0n 04 0
~ hhi hi hi i i h hia1-2h
0~ 0 M-60 4 0 0
$4 .A5 00 0 14 co0O 4~ 1 '34 ~cc. ~ cc
41 2 0) 414 q.41.cc wc wc cco 4)' .4 (
v- 1.4 054 u 01.4 0 w44 -A
Aj 0 0co .61 1.4 c 140 ca acc~ 0c co * ' 0 0 aw w sgo4
0 '. '44 3 aa 410a.. C0 0: (A cc (D00 0 0
0 0 0 04.4 w1r. 0 0 -4 0 u 4H1.
w 0 54 '4 0- AU us'
w4 1. 54 w4 -H 4 w 4. w 0 &a. C 0)411 c w V 0a0J 44 a3 w -Hr.r
1.4 A r 0 a a w0 140 000 o a "o1 0 tY H (' C4 on %w Vy4 r41 14 C" w4 0413 .,H4 cc'
C4
b-4
(A aa wA to"0o 0 0 0 u -r
E-0I - - 4 r- 9 00
w u u 0 0 00hUEUU4 0 U U. 0 0. wl w 0
0i m 0 0a 0 cc a
w0 0 0 0 r. o r0 0 0: 0 L 1.E-' ( 0 0 0 0 ) 54 w 4() N cu 0 l (a tOwc c ~ 21 3 o X w c co 00 0 0 0)-H 4) a 4 0)
03 a a a J CC-aj 0 QW 0 .4 u 0lH 1.4 5. 1. 5.4 $O 4 1.4 .0 0. 0. 54 w. 0A 0M
0 0 0 0 0 0 0 54 1.40 0 0w41. 5.4 k4 V.4 w. s.4 0, V-4 5 4w,. r4 4 H.4 u~ N w 0 ciH..4 0 .
4Z4 9 5z w mo ow Ss 9m00 0 U U
. M IT H H4 en H- H r4 enH H
C ~ ~ ~ r- v-4 H C ~ 0 -
1-
I-
Cu0rd M
00 r44U
4) 0 0 ii4
or cc C4 '-l 4k 4d Yk
o w m 4be be C 0 0 0
0 .0 .0 0
w -4 _ _ _I- 1 4 .0 0-P4 -H a)04Cu z u
r0 r 0 . v r. r. 4) Wr. r r0 0P 0H a a
0z 0 .4 r-4 w M
P-4 Cu4 Cu -4 C4 CuonC C
- _
aca a aco Go~ N N N 0%47%V-
06 6
N4 ,4 (4 C4 N C4
4 _ _1-4
II
TABLE 1-2. FORTRAN SOFTWARE FOR NRL PASSIVE MICROWAVEITEMPORAL CLUTTER PROJECT
PROGRAM PURPOSERADCAL Obtain calibration data
RADDAT Convert 1984 raw data into BrightnessTemperatures.
RDDAT Convert 1983 raw data into Brightnessi Temperatures
RADOUT DARPA statistical analysis, 1984 Data
RDOUT DARPA statistical analysis, 1983 Data
RADCO Calculate calibration coefficients forRADCAL data.
I RADADD Calculates spectral interval powerfor given range of FFT coefficients
RADAVG Calculates average sky and enclosureFFT or Temporal Structure spectrafrom individual data files on tape,
normalizes enclosure data, calculatessky contribution and plots usingCalComp graphics
RADCOR Corrects windowed FFT files by multi-plying normalization factors
RADNOR Calculates sky component of clutter
data from individual disk files
FFTEST Test FFT subroutine by using severalinput spectra with known FFT structures
TPREAD Copies clutter data files from tapeto disk
TPSTOR Copies clutter data files from diski to tape.
TRANS Reads mixed format tapes files fromRADDAT and generates formatted files
of 4096 samples.
4
U 1-5
SECTION 2 - MICROWAVE TEMPORAL CLUTTERDATA REDUCTION PROGRAMS
CSC has developed the data reduction programs that aided in Lnterpre-ting and evaluating the 37GHz microwave temporal clutter data. Thedata acquisition system obtains thp raw radiometric data from themicrowave sensor; and the reduction programs convert the raw data intobrightness temperature data, which are stored on magnetic tape files,and other output products: computer-generated graphic plots, lineprinter data listings, and disk files. The FORTRAN programs (RADCAL.
RADDAT and RDDAT) operate on the NRL DECsystem-10 computer and arediscussed below.
g2.1 RADCAL DATA REDUCTION PROGRAM
The RADCAL program reduces the laboratory and field experiment 37GHzradiometer calibration data, providing the statistical informationrepresenting instrumentation operational and gain setting characteristics.This information is required to calculate the count (raw data)-to-antennatemperature and antenna-to-brightness temperature conversion coefficients.
*RADCAL produces line printer listings and histograms of the raw analog-to-
digital (A/D) converter counts for the radiometer output. It alsogenerates averages and standard deviations of these A/D counts alongwith the thermistor measurements of ambient and instrumention referencetemperatures.
2.2 RADDAT DATA REDUCTION PROGRAM
The RADDAT program performs the actual conversion of A/D counts intoantenna temperatures and then into brightness temperatures for a user-specified set of data records from the 1984 observation data sets.RADDAT calculates record scene temperatures using both field experimentand laboratory calibration coefficients calculated in RADCAL. These
coefficients, obtained for the beginning and end of the specifiedrecord range, are linearly interpolated to obtain a time dependentequation for antenna temperature calculations. It is assumed thatbrightness temperature equaled the antenna temperature for the 1984data. Several output products are produced by the program, includinghistograms and computer-generated graphic plots of brightness tempera-tures for each data record, and a computer tape containing brightness
temperatures, record numbers and timekeeping information.
I
2-1
Mme
!I
2.3 RDDAT DATA REDUCTION PROGRAMIThe program RDDAT is similar to RADDAT with software changes to enablethe input of 1983 target data.
During the 1983 experiment period, the data acquisition system wasadjusted to sample and store the radiometer signal at twice the speci-fied sampling rate. The raw data tapes produced during this observ-ing period contained data records with individual samples spaced 0.020seconds apart instead of the specified rate of sampling every 0.040
I seconds. The RDDAT program was modified to input every other samplepoint to conform to the specified sampling rate. This modificationreduced the number of samples per record from 26 to 13, storingon tape one-half the expected number of brightness temperature valuesfor a given record range. Antenna-to-brightness temperature calibra-tion coefficients were calculated for the 1983 data, and used in timedependent linear regression equations to calculate brightness tempera-tures from the antenna temperature data. In all otiher aspects, RDDAIis identical to RADDAT.
IijI
II
I
I
SECTION 3. - MICROWAVE TEMPORAL CLUTTERDATA ANALYSIS PROGRAMS
Brightness temperature data, reduced from 37GHz microwave radiometerraw data tapes, have been analyzed using a series of computer programswritten by CSC. The brightness temperature data have been processedusing the computer programs RADOUT for 1984 experimental data; RDOUTfor 1983 experimental data; and one or more of the normalization andaveraging programs RADADD, RADAVG, RADCOR and RADNOR. These programsare discussed below.
3.1 RADOUT DATA ANALYSIS PROGRAM
The RADOUT analysis program performs the bulk of the microwave brightnesstemperature temporal clutter data analysis. RADOUT is a modifiedversion of PROOUT (Reference 2), and was written specifically for thetemporal clutter project. The operating characteristics of RADOUTIalong with detailed discussions of the four statistical analysis proce-dures are presented below.
Magnetic tape data files, generated by the RADDAT computer program, wereinput data for RADOUT. Any data sets containing a power of two (2n )
elements could be analyzed, but data sets which contained 4096 elementsor brightness temperature values were selected for the final data base.
- A list of the data sets in the data base is presented in Table 3-1.Each data set contained at least 4096 elements. If more than 4096elements were available, only the first 4096 elements were used foranalysis.
The subprogram DARPA is driver for the temporal clutter data analysis,processing the data sets and controlling access to the various statisticalanalysis modules. Parameters within DARPA must be set by the program userto define data set size, analysis procedure, and output products. Out-put products from RADOUT included disk and magnetic tape data files,line printer listings, and computer-generated graphic plots. The outputproducts produced depended on which analysis module was accessed.
3-1
3-
II
TABLE 3-1 BRIGHTNESS TEMPERATURE DATA BASE
jFOR MICROWAVE TEMPORAL CLUTTER ANALYSIS_ _ _ _ __ _ __ _ _ _ _ _ _DATE TAPE/FILE RECORD RANGEj TARGET FILENAME
November 1,1983 3/8 473-1052 Aspna±t 1INovember 1,1983 3/8 1167-1733 Enclosure #1-83November 2,1983 1/2 510-1088 Water #1November 2,1983 2/1 538-.1109 Water #2November 2,1983 3/1 810-1383 Water #3November 2,1983 4/1 500'1052 Wood #1November 2,1983 5/1 460-1011 Wood #2November 2,1983 6/1 432- 984 Wood #3I November 2,1983 2/1 1395-2057 Enclosure #6-83November 2,1983 4/1 1276-1841 Enclosure #7-83November 2,1983 6/1 1143-1721 Enclosure #8-83November 3,1983 1/1 1811-2393 Grass #1November 3,1983 2/1 670-1240 Grass #2November 3,1983 3/1 888-1565 Grass #3November 3,1983 1/1 2586-3155 Enclosure #3-83November 3,1983 2/1 1486-2096 Enclosure #4-83November 3,1983. 3/1 1826-2368 Enclosure #5-83.November 17,1983 3/1 608-1037 Asphalt #2November 17,1983 3/1 2388-3058 Enclosure #2-83February 27,1984 1/1 306- 506 Sky #lAFebruary 27,1984 1/1 507- 723 Sky #1BFebruary 27, 1984 1/1 851-1011 Enclosure #1February 27,1984 1/1 1040-1-200 Enclosure #2February 28,1984 2/2 720-1049 Sky #2February 28,1984 3/1 2272-2473 Sky #3February 28,1984 4/1 955-1115 Sky #4-1AFebruary 28,1984 4/1 1116-1276 Sky #4-lBFebruary 28,1984 4/1 1277-1495 Sky #4-iCFebruary 28,1984 4/1 1677-1837 Sky #4-2AFebruary 28,1984 4/1 1838-2052 Sky #4-2BFebruary 28,1984 4/1 2855-3025 Sky #4-3AFebruary 28,1984 4/1 3026-3200 Sky #4-3BFebruary 28,1984 4/1 3535-3715 Sky #4-4ATebruary 28,1984 4/1 3716-3906 Sky #4-4BFebruary 28,1984 2/2 300- 460 Enclosure #3February 28,1984 2/2 452- 612 Enclosure #7February 28,1984 3/1 2830-2990 Enclosure #4February 28,1984 3/1 2691-2851 Enclosure #8February 28,1984 4/1 202- 361 Enclosure #5February 28,1984 4/1 2475-2635 Enclosure #6
II
P 3-2
t .....-.
1 3.1.1 THE FAST FOURIER TRANSFORM ANALYSIS MODULE
IFourier Transformation statistical analysis incorporates a Fast FourierTransform (FFT) coded into the subroutines FFT and HANN. The FFT
jalgorithm was developed to analyze each 4096-element brightness tempera-ture data set, representing a time series of 163.84 seconds in duration.The FFT algorithm defines the spectrum of this time series, and issimilarly structured to the International Mathematical and StatisticalLibrary (IMSL) subroutine, FFTRC. (Reference 3).
The spectrum obtained from the FFT algorithm is representative of afinite section of a infinite time series. Analysis of the backgroundI scene based on this finite section introduces erroneous signals orspectral leakage into the FFT spectrum (Reference 4). The spectralleakage results from frequency discontinuities at the boundaries offinite time series. Several schemes have been implemented to filterand reduce the spectral leakage introduced into the data sets during theFFT calculations. These schemes include overlapping correlations and
I window weighting of the data.
Reduction of spectral leakage can be obtained by multiplying each tempera-ture series by a weighting or filter factor. Two types of filteringwere used for this project. The first is a window function called thecosine squared, raised cosine, or Hanning window function. The Hanningwindow function is defined by Harris (Reference 6)! (3-1)
4) (n) 0.5 1.0 - cos2n (3-1)
for n - 0, 1, 2 )
where W(n) - Hanning function for element n+l
n - individual element in data time series
N - number of elements in time series
The second window function, the Hamming window, does not reduce theboundary data points to zero as does the Hanning window function. TheHamming window function is defined by Harris (Reference 7):
CJ' (n) - 0.54 0 .46 [os (2n 7T1 (3-2)
I for n 0, 1, 2, 3...,N-1 and
where ' (n)-Haming function for element n+l andI n and N are as above.
1 3-3
I . ... - " l _ ,. .
When windowing of data is required, the brightness temperature seriesare multiplied by eithertd(n) ort' (n), producing the weighted bright-ness temperature T'B (n):
T'B (n) TB (n) x W(n) or (3-3)
T'B (n) TB (n) x (j(n) (3-4)
The overlapping correlation scheme is incorporated to recover time seriesboundary data lost during window filtering. This procedure is outlinedin Harris (Reference 5). The original brightness temperature timeseries was broken up into several user-designated overlapping subseries.The program user specifies in the subroutine HANN his choice of 0, 3,or 6 overlapping subseries. For example, if the original 4096-elementseries was divided into three overlapping subseries, the first subserieswould contain elemepts numbered 1-2048, the second subseries would con-
3 tain elements numbered 1025-3072, and the third would contain elementsnumbered 2049-4096.
The FFT spectrum is calculated for each overlapping T'B time seriesand then averaged. This averaged overlapping filtered power spectrum isplotted versus frequency after a normalization factor is applied toadjust the spectrum amplitude back to its original level (see Section 3.2).Figure 3-1 is an example of a averaged normalized overlapping HANN-filteredFFT power spectrum for a background target.
1 3.1.2 THE TEMPORAL STRUCTURE FUNCTION ANALYSIS MODULE
The Temporal Structure Function algorithms are executed by accessingthe subroutine TSFUNC. The Temporal Structure Function is a methodin which the intensity of baseline fluctuations or background cluttercan be calculated. The temporal structure DrB was calculated usingthe formula from Gagarin and Kutuza (Reference 8).
DTB ( t )-(TB (C) - TB (t-6t))2 (3-5)
where D (t temporal structure coefficient for
T (t) - brightness temperature at time t
TB( t+Ar) - brightness temperature at time t ."t
i and At- increment of time, seconds.
The temporal structure coefficients were calculated to obtain averagedsquare temperature differences for time differences ranging from 0.040seconds to 40.96 seconds. The Temporal Structure Function was thengraphically plotted versus time. Figure 3-2 shows the temporal structureof a typical background.
1 '2-4I .
-j
1 -1
ovC)
I 3-5
rKr
CD
-0 >U
CD LI
00A5c w
Li
3-6- Ir2~ ., -
3.1.3 TEMPERATURE AVERAGING AND DIFFERENCING ANALYSIS MODULE
Temperature averaging and differencing is a statistical method basedon an analysis presented by Gagarin and Kutuza (Reference 9). Thisanalysis is initiated by calculating the average brightness temperatureand standard deviation for the 4096-element data set. Also, an averagetemperature difference between adjacent elements and the associatedstandard deviation are calculated. The second step in this analysisaverages adjacent elements together to obtain a new data set comprisedof one-half the original number of elements. The averaging and differenc-ing calculations are then performed on the new data set. This schemeis repeated until less than four elements remain to be averaged together.At that point, the results is stored on disk file for line printeroutput. Table 3-2 is an example of a typical target printout of theII averaging and differencing analysis. As shown in the table, the in-tegration time is increased by averaging elements together. The effectof longer integration on the data statistical averages and standard
I deviations can be analyzed using this method.
3.1.4 AUTO-CORRELATION ANALYSIS MODULE
Auto-correlation is the fourth statistical algorithm in RADOUT accessed
by calling the subroutine AUTO. The auto-correlation analysis allows
the detection of a known signal in the presense of noise. The techni-
que also helps to analyze background clutter noise characteristics and
temporal variations. The auto-correlation algorithm was based on the
formula from Clay and Medwin (Reference 10):II N-k
Cxx(k) = 1 a XnXn +kN x n=O
I whereCxx(k) - auto-correlation coefficient
k - spacing parameter
x2 - variance of brightness temperatures
N - number of elements in data set
Xn - brightness temperature for element nIn + k - brightness temperature for element n+k
After the auto-correlation coefficients were calculated for values of kranging from 1 to N/4, the auto-correlation coefficients were graphicallyplotted versus time or k. Figure 3-3 represents an auto-correlation
analysis with a known signal present. The auto-correlation analysiswas not used to analyze the entire data base because it was so time-
consuming and costly.
33-7
I
TABLE 3-2
FRIUHTiES "fEi'PERATURE. AIAL..Y0Ia;
FUR SAM~PLE AVEIR:ArINC )AND r:F', D:LFFERENCNG
37H CHAINNEL 3/V CHAMNL.L
UF TIME ~AVG. Ac VOAGp ul-. . . . . . . . .. .- .v ... .."lAMOS' PERT;'-0 U;R'I" : S ft., ! 'EHi" 1.*'l-t0 .;f:" STD TOP " ii? STI:
( LC) I E.iiP DEV DIFF DE.V, Ei ii- DLv . DIF DE';
(°K) (OK) (°K) (OK) (°K) (°K) (°K) (,K)
0 4?6 O. 0"10 13b. U6 0,72 0.00 UU J.6., 0 1 2 0.00 2.26
20.4'I 01000 135.06 0 * U 0.00 0 , :? 1.,6 1,. O,J. -000 0Y.6
- 0 1420 .13 5 .136 0,4V 0.00 0 . .,. .... 0 * 1 --- 0. 00 0 74131 1 01320 13,, ., , 0,00 O,'i( .1J3 0,33 -0.00 0.3,
0"0 00 0.10 1. .1::: 0.1. -0.00 0.:36!"6 0.640 135,U6 0.12 0.00 Q . i"o .i.) Jo 3 .0091 0.00 ( '.11
,71 1,7 .20 13b. 06 0.10 0 1 01 0 . I:: 1. 6 1. :1 0. 08 0,00 0 1..
r.4 10.240 13., 04(U/ 0-,01 (.0 l:,3 0,05 -0.00 (.00
00 00.0 .0 1.; ii.x 0 .o.~ 0100 (J. 1),." 10 AD 0} 3 .. L,': ° . 0 , 0 ,0:. -...,1 0' :.,. ( 01 0.,0 I 0,10
3-8 .. no
I
ozrn
U 0 U3
I o
- I-
I U
c
!00
IIn
.4k
1 .3-9
II
II
3.2 RDOUT DATA ANALYSIS PROGRAM
The RDOUT computer program is a modified version of RADOUT; RDOUT wasdeveloped to accept data tapes generated from the RDDAT program. As
I stated in Section 2, all data collected during the 1983 experiments mustbe separately processed because there are 13 samples per record insteadof the standard 26 samples per record. All statistical analysis modulesin RDOUT are identical to those in the RADOUT modules.
3.3 AVERAGING AND NORMALIZATION PROGRAMS
The analysis programs (RADADD, RADAVG, RADCOR, and RADNOR) were writtenby CSC to analyze FFT and temporal structure data generated by theRADOUT and RDOUT programs. The Data Normalization, Averaging, andDifferencing (DNAD) procedure utilized these programs to perform thefollowing tasks:
o Average tbgether similar background target data sets
o Average horizontal (H) and vertical (V) polarization datatogether (1984 data only)
I o Normalize associated enclosure data
o Subtract normalized enclosure data from the target datasets to obtain a differenced target data set
Enclosure data were normalized to compensate for temperature variationsbetween the target scene and the enclosure absorber.
I The target or background scene data is comprised of two temporal cluttercomponents: the target temporal clutter and instrumental noise temporal
clutter. Instrumental noise temporal clutter is identified from the norma-lized enclosure data. Target temporal clutter can be calculated by subtract-ing the normalized enclosure data from the original target scene data.Target Scene data sets used in the DNAD procedure are listed in Table 3-3.Representative FFT spectra processed through DNAD are graphicallyillustrated in Figures 3-1, 3-4, and 3-5. Figure 3-1 illustrates aFFT spectrum of an average target scene which was fully processed en-
closure associated with the target scene Figure 3-1. Figure 3-5graphically displays the difference spectrum between Figure 3-1 and3-4. This FFT spectrum represents the target temporal clutter contri-bution of background noise.
The processing procedure for the temporal structure data was similarto that for the FFT data with an additional step. The target cluttercomponent temporal structure array was divided by the normalized tar-get clutter component temporal structure array. This additional steppinpoints the time lag necessary for background target noise to dominateinstrumentation noise. If the amplitude of the temporal structurecurve was greater than 1, target temporal clutter dominated. When thetemporal-structure curve amplitude was less than 1, instrumental noisewas dominant.
I l
The fully processed temporal structure arrays are graphically plottedversus time between samples. Figure 3-2 illustrates an averagedtarget scene temporal structure signature. Figure 3-6 illustratesthe enclosure temporal structure, and the difference or target componenttemporal structure signature is displayed in Figure 3-7. The temporalstructure for the normalized target clutter component is illustrated inFigure 3-8. As shown in Figure 3-8, target clutter or background noisedominate the measurement for sample lag times greater than approxiately
seconds. For lag times less than 3 seconds, instrumental noisedominates the target clutter.
[
[[I ,
I
i
II
III!!I
It TABLE 3-3
BRIGHTNESS TEMPERATURE DATA SETS USED FORTARGET AVERAGING
AVERAGE FILE NAME TARGETS AVERAGED
Asphalt Asphalt #1; November 1, 1983Asphalt #2; November 17, 1983Water Water #1, #2, 3; November 2, 1983Wood Wood #1, #2, #3; November 2, 1983
Grass Grass #1, #2, #3; November 3, 1983Enclosure83A Enclosure #1-83; November 1, 1983Enclosure #2-83; November 17, 1983Enclosure83B Enclosure #6-83; #7-83, #8-83November 2, 1983
Enclosure83C Enclosure #3-83, #4-83, #5-83;November 3, 1983Sky Feb.27 Sky #1lA, #lB; February 27, 1984Sky Feb.28A Sky #4-1A, #4-1B, #4-IC, #4-2A,#4-2B, #4-3A, #4-3B, #4-4A;February 28, 1984
Enclosure84A Enclosure #1, #2; February 27, 1984Enclosure84B Enclosure #1, #2; February 27, 1984Enclosure #3, #4, #5, #6, #7, #8;February 28, 1984
IIIIiII
3-12
IAL * #
I
I
I 0.,
04E-
I
-zz
I low 0IL
XI*M
3,1
3-].0
__________________
cno-
CA,
-
I fz. E-4
I 4
0=
0a
00
I 0a
4
lCD
I.:-
Li L iiCD
'D E-
3-15,
- w
m CA
'LL]7
I D QL
~awa
DC
r , I
II p
*f GD C,3
3-16
CD C
121W'
13, 06IwU-]w
00
Ic26i-I to
1 Cfl) 140
iLL))
U, L', PJ
N~~~~C S1JO9 40fJY9d~ P
9 .3-17
* IAPPENDIX A - SOFTWARE DOCUMENTATIONI
CSC has provided instruction on the use of the software developed forthe microwave temporal clutter project. This instruction included in-code program documentation and an orientation seminar held at the NRL.
In-code documentation was provided to enable the software user to followthe coded logic. The documentation consisted of prologue descriptionsand in-line comment statements. Prologue descriptions were written formain programs and subroutines. The prologues followed a standard formatand identified the code function, the accessed arguments and COMMONstatements, and the subprograms calling and being called. In-linecomment statements were written throughout the code to describe functionalareas of the program and explain the flow of specific algorithms.IIA software orientation seminar was held for I4RL personnel on August 6, 1984.CSC personnel provided instructions in software program execution anddiscussed specific aspects of all software developed for the temporalclutter project.
!II'I'
II
I-I _ , i = -- II III _ I l
II
I
REFERENCES
1. Computer Sciences Corporation, CSC/TR-83/6003, Support for theNaval Research Laboratory Environmental Passive Microwave RemoteSensina Program - Final Contractor Report, M. F. Hartman,IApril, 1983, pp. 4-1 to 4-4.
2. Computer Sciences Corporation, CSC/TM-84/6079, NRL RadiometerMode Aircraft Data Processing Guide, D. Niver, June 1984, pp. 4-1
Ito 4-3.
3. The Internal Mathematical and Statistical Library, Volume 2,Chapter F," Forecasting; Econometrics; Time Series; Transforms,"June 1982, pp. FFTRC-I.
4. F. J. Harris, "On the Use of Windows for Harmonic Analysis withthe Discrete Fourier Transform," Proceedings of the IEEE, Vol. 66,No. 1, January 1978, pp. 52
5. Ibid. pp. 56
6. Ibid. pp. 60
7. Ibid. pp. 62
8. S. P. Gagarin and B. G. Kutuza, "Influence of Fluctuations inAtmospheric Radio Thermal Emission on the Sensitivity of a RadioTelescope," Institute of Radio Engineering and Electronics,Academy of Sciences of the USSR, Vol. 19, No. 11, November 1976,
1pp. 138
9. S. P. Gagarin and B. G. Kutuza, "Influence of Sca Roughness andAtmospheric Inhomogeneities on Microwave Radiation of theAtmospheric - Ocean System," IEEE Journal of Oceanic Engineering,Vol. OE-8, No. 2, April 1983, pp. 65
10. C. S. Clay and H. Medwin, Acoustical Oceanography - Principlesand Applications, New York: John Wiley and Sons, Co., 1977, pp. 441.
R
IIR-
I!
IK 14~ d'~P - ... I.
-- a.
.9,
I,A, r
'C SEMISmmEEEEEEE~j
ii,