Nature & Wildlife


External Calibration of SIR-B Imagery with Area- Extended and Point Targets

External Calibration of SIR-B Imagery with Area- Extended and Point Targets
of 9

Please download to get full document.

View again

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
  IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. GE-24, NO. 4, JULY 1986 External Calibration of SIR-B Imagery with Area- Extended and Point Targets MYRON C. DOBSON, MEMBER, IEEE, FAWWAZ T. ULABY, FELLOW, IEEE, DAVID R. BRUNFELDT, MEMBER, IEEE, AND DANIEL N. HELD, MEMBER, IEEE Abstract-Data takes on two ascending orbits ofthe Shuttle ImagingRadar-B (SIR-B) over an agricultural test site in west-central Illinois were used to establish end-to-end transfer functions for conversion ofthe digital numbers on the 8-bit image to values of the radar back- scattering coefficient u0 (m2/m2) in dB. The transferfunction for each data take was defined by the SIR-B response to an array of six cali- brated point targets of known radar cross section (transponders) and to alarge number of area-extended targets also with known radar crosssection as measured by externally calibrated, truck-mounted scatter- ometers. The radar crosssection of eachtransponder at the SIR-B cen- ter frequency wasmeasured on anantennarange asa function of the local angle of incidence. Two truck-mounted scatterometers observed 20-80 agricultural fields daily at 1.6 GHz with HH-polarization and at azimuth viewing angles and incidence angles equivalent to those of the SIR-B. The form of the transfer function is completely defined by the SIR- B receiver and the incoherentaveraging procedure incorporated into productionof the standard SIR-B image product. Assuming thatthe processing properlyaccounts forthe antenna gain, all transferfunc- tioncoefficients are known except for the thermal noise powerand a system  constant that has been shown to vary as a functionof un- commanded changes in the effective SIR-B transmit power. For each orbital pass, the SIR-B thermal noise was estimated from surfaceareas expected to yield specular reflection, and the system  constant was determined for each area on the SIR-B image containing atargetof known radar cross section. Both thepoint targets and thearea-ex- tended targets were found to yield nearly identical results with a mean difference of approximately 0.1 dB. For a given date, the standard error ofthe estimate for the system  constant as derived by this method is found to vary from +0.85 dB to +1.35 dB. The interpass variance of the transfer functions was found to be related to theob- served variance of the effective SIR-B transmit power. Application ofthe system transfer functions to SIR-B imagery permitted realization of science objectives by allowing comparison of multidate imageryon a common basis. Five of the six transponders also operated as calibrated receivers. For each of six data takes, two ascending and four descending, thereceivers were distributed over an areaextending approximately20 km in both range andazimuth directions. Foreach SIR-B data take, each receiver recorded the time historyofavoltage proportional to the in- cident power density at the ground. The observed azimuth beam formappeared to be nominal with respect to specifications. Preliminary analysis ofthe range pattern, which could not be ascertained in a direct fashionwith statistical confidence, indicatesthatthepattern may beManuscript received December 13, 1985; revised February 7, 1986. This workwas supported by the Jet Propulsion Laboratoryunder Contracts 956921 and 957191 as subcontracts under NASA Contract NAS7-918. M. C. Dobson and F. T. Ulaby are with the Electrical Engineeringand Computer Science Department, University of Michigan, Ann Arbor, MI 48109. D.R. Brunfeldt is with the Applied Microwave Corporation, Lawrence, KS 66046. D. N. Held is with the Jet PropulsionLaboratory, California Institute of Technology, Pasadena, CA 91103. IEEE Log Number 8608572. nominalprovided thatthetrue antenna boresight is estimated to an accuracy of +20 via preliminary estimates ofthe STS ephemeris. Fi- nally, the uncommanded loss of effective transmitpower, which has been attributed to arcing, was found to average 7.1 dB and vary from pass-to-pass by approximately 3 dB. I. INTRODUCTION JN OCTOBER OF 1984, the Space Shuttle Challenger Icarried the Shuttle ImagingRadar (SIR-B) as part of the payloadon the STS-41G mission. The SIR-B instrument is an HH-polarized L-band SAR (1.28 GHz) capable of operating over incidenceangles from about 15° to 600 relative to nadir. Although the mission was plagued by a series of hardware malfunctions, avast quantity of data was acquiredover both water and land surfaces. Much of this data was digitally recorded and transferredvia TDRS- 1 for subsequent digital processing by the Jet Propulsion Laboratory. The digital SAR correlator incorporates both known and modeled system parameters in an attempt to produce digital image products free from system-related artifacts both in range and azimuth.Since many of thescience objectives of the mission re- quired relative calibration of the SIR-Bimagery in order to compare either multiangle or multisite observations,an important technical issue is the stability of the SIR-B in- strument and the adequacy of the antenna patterns as- sumed in the SAR correlator. In addition, it is desirable to evaluate existent means for providing end-to-end sys- tem transfer functions for the conversion of image digital numbers into units of radar backscattering coefficient a° (m2/M2). In order to address these technicalobjectives, a test site was established in west-central Illinoisat the intersect point ofprojected ascending anddescending SIR-B cov- erage. Preceding the ascending data takes,the Challenger generally conducted orbital alignmentmaneuvers so that the shuttle ephemerides couldbe precisely determined. During the mission, six digital data takes over this site were obtained from three azimuth view angles and with local angles ofincidence from - 17° to 590 as given in Table I. Of the six data takes, digital imagery was pro- duced for five swaths asillustrated in Fig. 1. On the ground, the testsite comprised an irregular area of roughly 250 km2 intended to transect bothascending and descending image swaths. For science objectives re- lated to the radar backscattering from vegetation canopies 0196-2892/86/0700-0453$01 .00   1986 IEEE 453  IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING,VOL. GE-24, NO. 4, JULY 1986 ILLINOIS SIR-B SITE Fig. 1. Map of SIR-B coverage. Each swath is annotatedwith its respec- tive data take number. TABLE I DATA TAKES OVER THE ILLINOIS TEST SITE Look Local Angle Patet Data-take Direction of Incidence 10/07/84 10/08/84 10/09/84 10/10/84 10/11/8438 *49,2 54 70.1 86  97.2 NE NW NE NE NE SW 17 3038 50 59 31 *indicates imaged swath contains fields co-observed by the truck-mounted scatterometers. and underlying soil, this area was both extensively and intensively monitored on a daily basis for soil moisture andcanopy biophysical conditions. For technical objec- tives,the site was instrumented with an array of six point targets and traversed on a daily basis by two L-band truck- mounted scatterometers. The point targets, active radar calibrators (ARC's), function both as calibrated receivers and as transponders. Each ARC consists of two patch antennas (transmit and receive) connected by a detection circuit and a high-gain RF amplifier [1]. As a receiver,interfacing the ARC with an instrumentation tape recorder permitted retrieval of the time-history of the SIR-B transmit power to yield azimuth cuts of the SIR-B antenna pattern. During each of the six SIR-B data takes, the ARC's were distributed over the test site with spatial separations of up to 20 km in range and/ or azimuth. Subsequent processing of this data using the ARC transfer functions and estimates of the shuttle posi- tion and velocity to calculate the SIR-Bpower density at the ground also provided a means for examination of the cross-track(range) antenna pattern and comparison of pass-to-pass variability in the SIR-B transmitter s output power. As transponders, the ARC's perform as point targets of known radar cross section. When properly deployed within an area characterized by a relatively low a0, the ARC is imaged as a bright target that serves as a calibra- tion reference for establishment of the image transfer function. Although six ARC's were deployed within the test site during each data take, misalignment of the SIR- B data window with respect to the test site on the descend- ing orbits (Fig. 1) restricted imaging of the ARC's to pri- marily the ascending data takes (DT 49.2 and 97.2). An additional method for computation of the image transfer functions was provided by the observation of many area-extended targets by the truck-mounted scatter- ometers. Two FM-CW scatterometers were used; both were operated at 1.6 GHz with HH-polarization. On a daily basis, each system observed 40 to 80 different ag- ricultural fields distributed along either north/south or east/west transects of the site. Scatterometer observations were made at angles ofincidence andazimuth view angles equivalent to those forthe SIR-B on any given orbit. The system output products were a0 as externally referenced to a Luneberg lens. Each a0 is the average of at least 100 independent samplesdepending upon angle of incidence. As a consequence, a statistically significant sample pop- ulation was created for establishing thetransfer function for each image that covered the area co-observed by the truck systems and SIR-B. II. SIR-B ANTENNA PATTERNS AND TRANSMIT POWER Foreach of the six SIR-B datatakes over the site, the ARC's wereused to record the time history of theincident power density. The outputvoltage of the ARC is propor- tional to the received power at the input to the ARC re- ceiver. The transfer function of each ARC was estab- lished in the laboratory by injecting signals of pulse width T = 30.3 ,us and monitoring the output voltage as a func- tion of input power Prand pulse repetition frequency (PRF) from 1200 to 2000 Hz VOut = K(PRF) x Pr  1) where K(PRF) is the ARC transfer coefficient. A. Azimuth Patterns An example of the time-history of an ARC output volt- age recorded during a SIR-B overpass on October 11, 1984 454  455 DOBSON etal.: EXTERNAL CALIBRATIONOF SIR B IMAGERY 0.125 - SIR-B AZIMUTHPATTERN ASSUMED SIR-B RANGE PATTERN 0.100 ID 0.1 01 cu 0 a) 0) 0.1 5 F Power 0.63 sec Orbit 97.210/11/84 ARC  1 20 PtGt = 86.72 dBm Range Displacement fromEstimated SIR-B Boresight = 3.97 km or -3.01° m c, .CIO u -10 dB Sidelobe 0 Time (Seconds) Fig. 2. Typical azimuth beam-form observed by an ARC. TABLE II AZIMUTH ANTENNA PATTERN SUMMARY Standard N of 1ean Deviation Samples Beamwidth 3dB 1 095 038' 26 Nul Ito Null 2.421°.049° 20 Sidelobe Level: Lead -1 1.96 dB 0.73 dB 20 Lag -10.34 dB 0.74 dB 21 Values are basedupon preliminary ephermeris data. is shown in Fig. 2. Depending on the level of the ouptut power of the SIR-B transmitter, the ARC receivers gen- erally recorded the second or third sidelobes within thelinear portion of the receiver response. The average ob- served sidelobe levels are given in Table II as   2.0 and   10.3 dB for the lead and lag sidelobes (relative to STS attitude), respectively. The half-power (3 dB) and null- to-null beamwidths are calculated from estimates of the STS velocity and range to target as derived from the pre- liminary ephemeris data. The average preliminary esti- mate of 183 dB iS 1. 1. The variances around the calculated means are found to be very small. The azimuth cuts recorded by the ARC's exhibit no evi- dence of pulse-to-pulse variation in SIR-B transmit power over their respective 3-4 s time histories   = 5 000 pulses). All measured patterns are smooth and well behaved. B. Range Patterns The range antenna pattern assumed by JPL in the SIR- B digital correlator is shown in Fig. 3. The research ob- Angle OffBoresight(Degrees) Fig. 3. Assumed model of SIR-B range antenna pattern used in processor. SIR-B ANTENNA PATTERNS FROM ARCs Image   Swat *.-.Active Radar Calibrator L ARC Beam Center (Boresight) Fig. 4. Sketch of active radar calibrator ARC deployment. jective was to use the ARC measurements to verify this model for the antenna as deployed in orbit. Since the range (elevation) pattern of the SIR-B antenna cannot be measured directly from the observations made by a single ARC, it must be inferred from the observa- tions of many receivers distributed in the range direction. The deployment of the ARC's with respect to any given shuttleorbit is depictedschematically in Fig. 4. Typi- cally, the ARC's were distributed over a 15 to 20 km range extent, which roughly corresponds to 50 in angle depend- ing upon the slant range for a particular orbit. As a con-  IEEE TRANSACTIONS ON GEOSCIENCE ANDREMOTE SENSING, VOL. GE-24, NO. 4, JULY 1986 DIFFERENCE IN BORESIGHT POSITIONAS ance in SIR-B pass-to-pass transmit power as indicated by MEASURED BY ARCS AND ASSUMED FROM the peaks of the curves plotted in Fig. 5 and 2) the esti- mates of the antenna boresight derived from the prelimi- nary ephemerides can lead to as much as a 3° error in the Data-Take Line/Symbol application of the antenna pattern to the radar data in the 97.2 86.1 digital correlator. The angular offsets depicted in Fig.5 70.1 a are believed to be the result of two factors: 1) errors in 90 ~~~~~~~~~~~~54.1 9 -549.2 the preliminary ephemerides and 2) discrepancies be- 89 38.1 * tween the true antenna attitude as deployed in the shuttle bay and the shuttleattitudeas measured in the noseof the 88 1 shuttle. The first factor canbe tested and may be correct- able by using ephemerides based upon the shuttle  path 87 _-97.2 ,/ >d/ <86.1 ape provided by the Johnson Space Center six months after the mission. The second factor is related to attitude and location-dependent thermal loading on the shuttle 850 / X X 54.1 which, at present,yields a noncorrectable error source withan uncertainty estimated to +10. 84 - /O / \\\\\XN Hence, while the true range pattern of the SIR-B an- tenna cannot be defined from the ARC measurements, it 942 is shown that errors in the preliminary ephemerides can 812 _ / \ lead to substantial misapplication of the modeled antenna 38.1 gainfunction indigital processing of the SIR-B data if the 81 - \assumedmodel is correct. The expected magnitudes of L0 *\ these errors are shown graphically in Fig.5 and listed in 4 -3 -2 -1 0   2 3 4 5 Table III. Conversely, the ARC data does not indicate Angular Offset (Degrees) that the assumed model for the rangegain pattern is in- Gt as observed by the ARCs for each data take. Angular offset correct. By assuming the true boresight location errors to lated from estimates of boresight location based upon preliminary be as indicated in Table III for each orbit and assuming rides. G, to betime constant (wherein only P, varies from pass- to-pass), then normalizing all ARC data with respect to ,e, it was possible to examine only a limited por- the maximum P, G, and boresight angle error for each pass the range pattern by forming a composite of the yields the data composite shown in Fig. 6. This approach, easurements for a given orbital pass. Ideally, many while certainly less than rigorous, indicates that the as- ,RC's or other receivers should be distributed over sumed beam form could be quite accurate. a greater range extent in order to clearly definethe shape of thepattern. Because the transmit power from SIR-B varied with time (probably due to arcing in a coaxial cable), the dif- ferences in incident power density recorded at the ground by the ARC's were not solely related to the range antenna pattern. Hence, the ARC output voltage was used toes- timate the product of the transmit power P, and the an- tenna gain G, by PtG, = 47rR2Pr/Aeff (2) where R is the slant range (as estimated at the stated angle of the antenna boresight relative to nadir using the prelim- inary ephemerides and assuming a spherical earth) and Aeff is the effective appertureof the ARC (as measured for each ARC on anantenna range). In Fig. 5, Pt G, is plotted for each ARC measurement as a function of the angular offset between the position of the ARC and the ground intersect of theantenna boresight as calculated from the preliminary ephemerides.The least squares fitting of the assumed beam pattern (as shown in Fig. 3) to the measured data yields the curves plotted in Fig. 5. If it is assumed that the modeled range pattern is correct, then it is clear that: 1) there is a significant vari- C. Transmit Power For the six orbits observed, the maximum P, G, esti- mated from the ARC's varied from 85.1 to 88.0 dBm (Ta- ble III). Since it is reasonable to assume that G, was time- constant, P, was found to vary over a 3-dB range. Given a nominal transmit power of 1 Kw (60 dBm) and a max- imum G, of 33.8 dB (from the antenna manufacturer and thermal vacuum tests), the maximum P,G, was expected to be 93.8 dBm as compared to the average observed (of six passes) maximum Pt G, of86.7 dBm. Hence, the aver- age observed loss in the expected transmit power is esti- mated as 7.1 dB with a pass-to-pass standarddeviation of 1.1 dB. The within-pass variability of SIR-B transmit power can be examined from plots of cross-track (range) averages of digital number asa function of time. For the Illinois data, these averages were found to be relatively constant yet they typically exhibited a short term fluctuation about the mean of +0.4 to 0.5 dB and a periodof about0.25 to 0.33 s. However, since the test site is a mixed agricultural scene with scattered towns, these variations may be scene related and not necessarily due to fluctuations in SIR-B transmit power. A sample is shown in Fig. 7 for datatake Fig. 5. P, is calcul ephemej sequenc tion of ARC m more A 456 co g  DOBSON etal.: EXTERNAL CALIBRATIONOF SIR-B IMAGERY TABLE III ESTIMATES OF MAXIMUM P, G, AND PRELIMINARYBORESIGHT ERRORS FROM RECEIVER MEASUREMENTS EstimatedDifferencebetween Range to Preliminary Estimates of Maximum Data-Take Elevation DataWindow Boresight Location and MaximumPower Angle Center that Predicted from Pt Gt Density (degrees)  Kmi) ARCs dBm (dB/m2) 38.116.40240 1.100 85.39-33.20 desc. (N) 49.2290 267 0.100 85.08 -34.45 asc. (N) 54.1 340282 0.95086.53-33.47 desc. (N) 70.1 480 353 -0.600* 88.01 -33.94 desc. (N) 86.1 560 425 2.820 87.49 -36.06 desc. (N) 97.2300 266 -1.450* 87.51 -31.98 asc. (S) * - ARC estimate is closer to nadir than preliminary prediction. 3 2Go 1 C -1 -2 -3 -4-5 ARCMEASUREMENTS RELATIVE 7 TO RANGE PATTERN 200- a 3 z -4 ._ bO .1_ bO 150- 1007 50- -5 -4 -3 -2 -1 0 1 2 3 Angle Off Boresight (Degrees) 4 5 Fig. 6. ARC measurements normalized to the antenna beam form as de- picted in Fig. 5 for each data take. 97.2. It is planned to iterate this analysis considering only scene elements for a single land-cover category (such as corn fields) once the SIR-B data is fully registered to ex- tensive ancillary data pertaining to the along-track crop- type distribution. As mentioned previously, no random pulse-to-pulse variation in P, was recorded by the ARC receivers. III. SIR-B IMAGE CALIBRATION Many potential applications of orbital imaging radar in- volving quantitative parameter retrievals require strict SIR-B 97.2 u   . 1000. 2000.3000. 4000.5000.6000.7000. Azimuth (Col) Fig. 7. Range-averaged digital number as afunctionof azimuth location for datatake97.2. The mean DN = 92.13 ± 5.42. standards of relative, if not absolute, radiometric calibra- tion of the imagery. Effective methods of image calibra- tion enable the comparison ofdataderivedover time, be- tween sites, and among suites of sensors using generalizedalgorithms. Considerable effort was taken to examine two external calibration methods to provide end-to-end trans- fer functions forthe SIR-B imagery obtainedover the Il- linois testsite. Targets of known radar cross section were provided by the ARC's (as transponders) and by area-ex- tended agricultural fields observed by two truck-mounted scatterometers. Application of these methods to the SIR-B data was partially frustrated by the fact that many passesdid not image the portion of the test site containing targets of V 1 Z ;u 4 457
Related Documents
View more
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks