Yinghong Sun's Intern Report
Summer Internship
at
Alaska Satellite Facility
by
Yinghong Sun
Electrical Engineering Department University of Alaska, Fairbanks
May 9th to September 4th, 1996
Table of Contents
- PREFACE
- INTRODUCTION
- APPENDIX A LITERATURE
- APPENDIX B ACRONYMS
PREFACE
There is a saying attributed to an ancient Chinese scholar that goes something like this:
"I hear, and I forget.
I see, and I remember.
I do, and I understand."
Most of us learn by a combination of audio, visual and kinetic activities. For my major subject "Satellite Communications", instead of having myself only "see" and "hear", I tried to find a chance to also "do and understand".
It was fortunate for me. This summer I got the chance to work as a summer intern at Alaska SAR Facility, which brought me four months fascinating life while working for those high quality satellite image products.
I want to say thanks to all the people working at ASF, who supported me during my internship. To my supervisor Jason Williams, who provided me the chance to come to the ASF and gave me the biggest help during the fascinating four months. To Parker Martyn, who helped me a lot during my distributed target calibration task and provided me the way to gather all the weather information for target analysis. To Tom George, who helped me to fly over Fairbanks area and got those beautiful aerial photographs for most of the targets. And of course to the other three interns, they are: Ty Sullins, Alexander Golitschek and Ylva Timner, they gave me such a great time at work and during all our trips.
[return to top]INTRODUCTION
The Alaska SAR Facility (ASF) is a sub-branch of the Geophysical Institute at the University of Alaska Fairbanks. ASF receives data from satellites orbiting in space and creates pictures which are distributed to users of the data. After processing and calibrating, the images are available for further observations. Scientists use these images to observe forest fires, sea ice, glacier movements, volcanoes etc.
The ASF Calval (Calibration and Validation) team calibrates and validates ASF SAR (Synthetic Aperture Radar) data products. When a satellite is first launched, they calibrate the SAR processor using targets. After a new satellite is calibrated, the Calval team routinely looks at SAR images from the satellite and make sure it is still calibrated. This is the "validation" part of their task. This involves repeating the same measurements with targets that they did when they calibrated the data.
As a summer intern at the ASF Calibration office, I worked with senior calibration engineer Jason Williams and his partner Parker Martyn. My intern project is called "Distributed Target Calibration Project", which is to calibrate the ERS-2 (European Remote Sensing Satellite) SAR data with a new calibration method by employing a series of natural targets instead of reflectors and transponders. Another main task for me was to upgrade a program called "verify", which is a C program used by calibration office to verify the data from satellites. The program reads in CEOS (Committee on Earth Observing Satellites) data files which are delivered with each ASF image product and then compares the values with the theoretical values. It can also be used to verify a new processor or any upgrading to an existing SAR processor. Besides these two tasks, I also had some other minor tasks, they are: support aerial photography planning, Toolik Lake corner reflector maintenance trip, and Delta Junction corner reflector maintenance trips.
[return to top]DISTRIBUTED TARGET CALIBRATION TASK
"Distributed target Calibration refers to external calibration using natural targets of large areas with homogeneous backscattering properties. One important benefit of using distributed calibration targets is that they measure the radar performance at various operating points within the system dynamic range. A second important advantage of distributed calibration sites is that they can be used as a direct measure of the cross-track variation in the received signal power as reflected in the digitized raw video signal after range compression." [1]
Compared to the point target calibration by using corner reflectors and transponders, another obvious benefit of using distributed targets is that there is no need of construction or deployment for the targets. And there is no need for power and safety consideration. That decreases the filed work to the lowest level.
[return to top]Distributed Target Calibration Processing
The calval department is currently calibrating ERS-2 using the Gen. III corner reflector array, which is made up of 9 corner reflectors (point targets). We can also use distributed targets. In this case, we measure the Back Scattering Coefficient ([sigma][omicron]) of the distributed target images and compare it to the actual [sigma][omicron].
The following graphic shows the flowchart of ERS-2 Calibration Processing by using both point target calibration and distributed target calibration.
The flowchart above shows that both Point Target Calibration and Distributed Target Calibration have the same algorithm. But unlike the actual RCS (Radar Cross Section) for a corner reflector, the actual [sigma][omicron] for a distributed target is very hard to accurately predict. However, since ERS-1 and ERS-2 are flying in tandem and ERS-1 has already been calibrated, we can use ERS-1 to measure the [sigma][omicron] (Back Scatter Coefficient) for our targets and use it as standard data to calibrate the ERS-2 satellite.
[return to top]DISTRIBUTED TARGET ANALYSIS
Before we start the analysis on the distributed targets, the following information needs to be collected:
a. SAR images over selected area
b. weather information for selected image pairs
c.aerial photographs or field photographs for targets
Then we can start our work. The figure below shows the way how I perform the analysis on the distributed targets.
Figure2.2 Distributed Target Analysis Flowchart
The three major steps for analysis are:
a. Target Selection
b.Image Pairs Selection
c. Poly Software Analysis
[return to top]TARGET SELECTON
The distributed target analysis starts with the selection on the desired distributed targets. Any target can be selected if they match the following requirements:
a. Homogeneous
b. Well characterized scattering properties
c. Consistent terrain properties which include: surface roughness property dielectric property
d. Easy to identify from SAR images
e. Large size
Homogeneous property (i.e., uniform in density/properties), is the most important characteristic for selected targets. The reason is: for a homogeneous scene, the exponentially distributed intensities of the scene elements will have constant mean, that will make the backscatter coefficient [sigma][omicron] to be constant. All of this provides the possibility to analysis the satellite images based on the backscatter coefficient [sigma][omicron].
Consistent terrain property is also important for targets selection. The terrain properties, such as surface roughness (rms height), dielectric constant (permittivity) and local slope, will change the characteristics of the reflected wave (amplitude, phase, and polarization).
The SAR image on the next page shows all the targets I selected for my analysis. These targets are located at Delta Junction area and can be described as following:
DS_1: Burned Area near Gerstle River Bison Range
DS_2&3: Farm Land (Black Spruce Forest) near Panaramic Bison Range
DS_4,5,6,7,8: Farmland (Brushes and Gravel)
DS_9: Moist Tundra Area at Donnelly Dome
DS_10,12,14: Forest along Tanana River
DS_15: Farmland on Buffalo Drop Zone near Air Force Base
Figure 2.3 Selected Distributed Targets on SAR Image
Next step, we need to determine the vegetation type of the selected targets. To do this, we will need either aerial photographs or the field photographs. The following four aerial photographs have been used to determine the vegetation type of the selected targets.
Figure 2.4 Aerial Photograph of Target #1
Figure 2.5 Aerial Photograph of Target #3
Figure 2.6 Aerial Photograph of Target #14
Figure 2.7 Aerial Photograph of Target #15
[return to top]IMAGE PAIRS SELECTION
Next step on analysis is to select image pairs between ERS-1 and ERS-2. The selected image pairs should satisfy the following requirements:
a. 24 Hours Time Difference
b. At same imaging geometry
c. No change in weather
It's easy to understand the reason for the first two conditions, however, the last condition is very important for us. Let's look at the following example.
The Table below shows weather information for Delta Junciton area from May 2,1996 to May 6,1996.
Table 2.1 Weather Information for Delta Junction Area
We notice that there was almost no weather change between May 2 and May 3, but a great change between May 5 and May 6. The rain at May 5 causes the precipitation change about 0.06 inch and relative humility change at 22%. Then what happened to our data from ERS-1 and ERS-2 satellites? Let's look at the following several figures.
Figure 2.8 [sigma][omicron] Change with RH for ERS-2
Figure 2.9 [sigma][omicron] Change with RH for ERS-1
Firgure 2.10 [Delta][sigma][omicron](ERD1 - ERS2) Change with RH
The three figures above show that the [sigma][omicron] will change greatly with the change of relative humility and precipitation. Also, for different type of targets, the change will be quite different. The Table below summarizes these results.
Table 2.2 Results from ERS-1 and ERS-2
Based on the weather information I got from Fort Greely Central Met Observatory, I chose the following ten image pairs for my distributed target analsysis.
Table 2.3 Selected Image Pairs Information
POLY ANALYSIS SOFTWARE
Poly is a program written by Alexander Golitschek to analyze the backscattering properties of area within SAR images. The program consists of functions for reading and creating files, calculating, and switch retrieving.
An important use of Poly is to calculate the backscatter coefficient [sigma][omicron], which is determined by using the following formula
[sigma][omicron]= 10*log10[a2*(DN2-a1*n(r))+a3]
where: [sigma][omicron] is the normalized radar cross section in dB a1 is the noise scale factor obtained from image leader file a2 is the linear conversion factor obtained from image leader file
a3 is the offset conversion factor obtained from image leader file DN is the noise scale factor obtained from image leader file n(r) is the noise of range at center obtained from image leader file Finally Poly will generate an ASCII output of the statistics of the image on screen as well as in a file.
[return to top]RESULTS OF ANALYSIS
The following results are obtained by using Microsoft Excel 5.0.
Result 1:
Figure 2.11 Comparison between E1 and E2 vs. Target Sequence
Comparing the image pairs between ERS-1 and ERS-2, we will notice that the ERS-2 image is always darker than ERS-1 image, which means the back scatter coefficient of ERS-2 is always lower than that of ERS-1. The same result can also be concluded from the following bar chart, and we will notice that the offset between the two satellites almost remain constant
while referring to different image pairs. Figure 2.12 [Delta][sigma][omicron](E1-E2) vs. Image Pair in Sequence
Result 2:
Figure 2.13 shows the second result from the target analysis. For different seasons, the offset between ERS-1 and ERS-2 will change. In winter (November - February), the offset is greater than in summer (March - October). The average offset value for winter is 4.9581 dB while it's only 4.4738 dB for summer.
Result 3:
Figure 2.14 [Delta][sigma][omicron](E1-E2) vs. Target in Sequence for target #15, a brush area near an air force base, the offset is quite higher than all the other targets.
[return to top]VERIFY UPGRADE TASK
ABOUT VERIFY
The verify program is used to support the work of the Calibration department, both in its daily work validating the ASF SAR data output and in the validation of the processor itself. It reads in a Verify Parameter File (VPF) and uses the information in the VPF to check parameters in the CEOS (Committee on Earth Observation Satellites) metadata that accompanies every ASF SAR products. It's currently only used on the conan workstation.
[return to top]UPGRADE TASK
The first version of verify, which is verify1.0, was written by Siegfried Jauch, who was a summer intern at the Calibration department during 1995. The verify1.0 can validate the old CEOS format metadata and can be used to check 47 parameters.
[return to top]THE OLD CEOS DATA FORMAT
The old CEOS data set consists of several files created by the SAR processor. These files are: SAR leader file(.ldr), SAR trailer file(.tlr), and SAR data file(.dat) file.
The leader and trailer files contain the following records:
| a. | volume descriptor record or file descriptor record | VDR |
| b. | data set summary record | DSS |
| c. | map projection data record | MAP |
| d. | platform position record | PPR |
| e. | attitude data record | ADR |
| f. | radiometric data record | RDR |
| g. | radiometric compensation record | RCR |
| h. | data quality summary record | DQS |
| i. | data histogram record | DHR |
| j. | range spectra record | RSR |
| k. | detailed processing parameters record | DPR |
| l. | calibration data record | CDR |
| m. | facility related record | FRR |
A record can be found either in both the leader and the trailer file or only in one of them. Every record start with 12 binary header bytes, which includes the record identification code and the length of the record. The rest of the record includes locations of the values and strings of the given information, implemented as ASCII characters.
[return to top]THE NEW CEOS DATA FORMAT
The new CEOS format data set only consists of two files. They are: SAR leader file (.L) and SAR data file (.D). The data file still contains the value of each pixel of the image, but the header has increased form 12 bytes to 192 bytes.
The new CEOS data products include both CEOS image products and CEOS signal data product.
The CEOS Image Products are:
GRF: Geo-referenced (Ground range or slant range)
GCD: Geocoded
GTC: Geocoded and Terrain Corrected
The CEOS Signal Data Product is:
CSD: Computer Compatible Signal Data
The following table is a summary of the ASF CEOS Products and CEOS format. B Table 3.1 ASF CEOS Products and CEOS Format
[return to top]
IMPLEMENTAION
The verify2 command line:
verify2 infile [-v] [-n] [-e2|-j1|-r1] [-a] [-c cal_file]
infile Input file name with CEOS data -v verbose, all information is print on the screen. Default only
prints the non-valid information
| -n | Report file will not be generated |
| Default generates the report file | |
| -e1 | ERS-1 <Default> |
| -e2 | ERS-2 |
| -j1 | JERS-1 |
| -r1 | RADARSAT |
| -a | ascending, <default is descending> |
| -c | read in the name of the calibration file |
| cal_file | instead of using the implemented validation files. |
| The path must be included. |
The "verify2" program analyzes the command line and finds the needed files. Then it starts to scan the leader file for the records and their locations. The main loop takes a line out of the appropriate validation file and determines the actual value of the dataset. It validates the value by comparing it with the theoretical value, string, or range in the validation file.
After the validation of the values, verify creates a histogram for the datafile if the file is there. It compares the values of the histogram with the values given in the DHR record and prints the result on the screen. If the result is negative and the [-n] option is not chosen you can find both histograms in the infilename.verify file but not on the screen.
[return to top]FLOW CHART OF VERIFY2 MAIN
[return to top]
THE VALIDATION FILES
Verify2 has several implemented validation files. The allowed combinations of datatype and matchtype are:
INTEGER EXACT
STRING STRING
DOUBLE RANGE
The validation files are ASCII files and the user can edit them to adapt the files to his demands.
At present, the following validation files are implemented:
validation_file_ERS1_ascending
validation_file_ERS1_ascending_HR
validation_file_ERS1_descending
validation_file_ERS1_descending_HR
validation_file_JERS1_ascending_HR
validation_file_JERS1_descending_HR
validation_file_RADARSAT1_ascending
validation_file_RADARSAT1_ascending_HR
validation_file_RADARSAT1_descending
validation_file_RADARSAT1_descending_HR
Further work are is needed on the validation file series for ERS-2 and on low resolution images for JERS-1.
[return to top]MINOR TASKS
[return to top]TOOLIK LAKE CORNER REFLECTOR MAINTENANCE TRIP
Toolik Lake is located 160 km to the south of Prudhoe Bay, the oil town on the Arctic Ocean in the north of Alaska. It is currently a research camp for Arctic Biology at University of Alaska.
The Toolik Lake maintenance trip was scheduled from the 25th to the 28th of May. It's the first time for me to take such a long trip by car - 15 hours driving - and it was also the first time I worked more than 18 hours per day. I was really tired after the four-day trip. But it was a great trip!
Most of the time during the trip was spent maintaining and setting up of the reflectors. Maintaining the corner reflectors at the top of a mountain was hard work. We had to trek up a mountain to the reflector over wet tundra partially covered by snow. Some of the reflectors were seriously bent by the snow and ice. The front edge levels of the reflectors varied between 0.6 to 6.0 degree off horizontal. That was terrible! We removed most of the ice and snow, but the reflector were still bent. There will be a lot of work to do if the reflectors are to be completely repaired.
Setting up new reflectors was hard work for us. The first time we set up a new reflector was during the second day of our trip. After spending all day maintaining the old reflectors, we began setting up new reflectors at 10:00pm and did not finish work until 1:30 am. It was very cold outside, and we had to wear lots of warm clothing.
Although we were working hard during those days, we still had lots of fun. There were so many interesting sightings on our trip: animals, mountains, rivers and oceans. I loved the Yukon River and Arctic Ocean so much that I took most of my pictures at these two sites. But unfortunately, I forgot to use the flash on my camera at Yukon River and all my pictures from there turned out very dark. The only thing that could be identified from those photos was the moon.
[return to top]DELTA JUNCTION MAINTENANCE TRIP
Besides the four day maintenance trip to Toolik Lake, we also took a lot of trips to Delta Junction to maintain the Delta Corner Reflector Array. These trips were part of the RADARSAT calibration program at ASF. The reflector array we used at Delta Junction is:
APPENDIX A LITERATURE
[1] Curlander, McDonough: "Synthetic Aperture Radar", Wiley-Interscience 1991
[2] Jason Williams: "Official Calibration Plan V1.0", ASF 1995
[3] JPL: "ASF Product Specification V1.0", December 4, 1995
[4] Siegfried Jauch: " Technical Report on Verify", summer, 1995
[return to top]APPENDIX B ACRONYMS
| ASF | Alaska SAR Facility |
| CalVal | Calibration/Validation Group |
| CEOS | Committee on Earth Observing Satellites |
| JPL | Jet Propulsion Laboratory |
| SAR | Synthetic Aperture Radar |
| ERS1 | European Remote Sensing Satellite 1 |
| ERS2 | European Remote Sensing Satellite 2 |
| JERS | Japanese Earth Resources Satellite |
| RADARSAT | Canadian SAR Satellite |
| RCS | Radar Cross Section |
| [sigma][omicron] | Back Scatter Coefficient |
| RH | Relative Humility |
VPF Verify Parameter File
[return to top]