Geometric Error Budget Analysis for TerraSAR-X
Contact Persons
Dipl. Ing. ETH Othmar Frey |
Keywords
SAR, TerraSAR-X, Geometric Error Budget Analysis, Geocoding, Simulation
Abstract
TerraSAR-X is a public-private-partnership project by the German Aerospace Center (DLR) and Infoterra, which is a subsidiary of the aerospace company Astrium. The spaceborne synthetic aperture radar system TerraSAR-X is intended to produce high resolution radar images for a broad range of applications and purposes. To meet the different requirements regarding the spatial resolution TerraSAR-X is designed to operate in three modes, Stripmap, ScanSAR and Spotlight mode, where the latter should provide image data suitable for image products of close to 1m resolution and a pixel localization accuracy of a couple of meters.
On behalf of the German Aerospace Center (DLR / DFD) a geometric error budget analysis regarding the high spatial resolution of the future TerraSAR-X system has been carried out. The following three work packages had been defined between the German Aerospace Center and Remote Sensing Laboratories RSL, University of Zurich:
- Analysis of the imaging model regarding the high spatial resolution and the ScanSAR mode.
- Definition of requirements for a high accuracy terrain geocoding of Spotlight mode data.
- Investigation of a possible improvement of the range dependent antenna gain pattern correction using a digital terrain model (DTM).
- Implementation of an Atmospheric Path Delay Model.
The higher the quality requirements the more crucial get errors in the parameters of the processing and post processing procedure. This study focuses on post processing, namely forward and backward geocoding.
Within the first two work packages, the impact of the following error sources on geocoded products has been investigated:
- orbit position
- datum shift parameters
- cartography (ground control points)
- geoid, digital terrain model (DTM), difference between DTM and true earth's surface
- atmospheric refraction
- sampling window start time (SWST)
- Doppler frequency
In order to ease the analysis and provide reusability of the analysis tools an Error Budget Analysis Tool "EBAT" has been developed in MatlabTM. EBAT comes with a graphical user interface. It can simulate forward and backward geocoding for all error sources mentioned above.
The results can be plotted and/or saved to an ASCII file having a consistent format and layout. (E.g. automatically generated caption). The following figure shows a screen shot of the program's graphical user interface after the calculation step, ready to plot and/or save the results.
The figure below represents the resulting plot when the 'Plot' Button is clicked using the parameter specifications of the GUI screen shot above.
The third work package aims at an optimized radiometric correction of SAR images. Within most SAR processing systems radiometric distortions due to the elevation antenna gain pattern are corrected using a range dependent function. However, the amount of the radiometric distortion of a back scatterer's representation in the image is determined by the geometric constellation between the antenna's position and attitude, and the position of the radar back scatterer.
In mountainous regions, situations occur, where two back scatterers lie within the same range distance to the sensor, although they have completely different coordinates. Hence, the elevation angle differs considerably, which implies a varying radiometric distortion in the image. If a range dependent correction function was applied, both back scatterers would be corrected by the same amount, although they experience different distortions (see figure below). Within this work package the effect of terrain height differences within a SAR scene on the radiometric distortion has been investigated.
Within the last work package the goal is to correct the path delay of electromagnetic waves traveling through the atmosphere. Path delays caused especially by ionosphere and troposphere can amount up to 3 m for TerraSAR-X (one way, zenithal). The model estimates the total path delay using only date, time and scene location. The so calculated path delay can be improved by optional input of measured data (e.g. water vapour, total electron content TEC or air pressure).
Project Status
The first phase of the project started in June 2002 and was successfully completed by the end of February 2003. The second phase started in September 2003 and was completed in early 2004.
Some Results
Actual pixel localization errors for Spotlight mode data will probably be at least a factor of 5-10 bigger compared to a preliminary estimation, which assumed ideal (i.e. error free) conditions regarding the DTM and the tie points. In general, the accuracy requirements concerning the digital terrain models seem to pose the biggest problems. It can not be assumed that DTMs of sufficient quality will be available world-wide. In addition, the required quality of maps for extraction of ground control points is not met for many areas in the world. Global geoid models are of insufficient quality. Vegetation covered and built-up areas degrade the achievable accuracy (available digital surface models (SRTM) might be not accurate enough). The influence of degradation due to datum shift errors heavily depends on the availability of new, precisely defined reference frames. Atmospheric refraction should be considered by simple modeling. Depending on the size of these error budgets for a particular situation the ScanSAR mode pixel localization requirements will be met or not. For areas, where no auxiliary data of good quality is available, actual errors will surely exceed the imposed requirements.
The radiometric accuracy requirements are not met even for rather moderate topography. The absolute radiometric accuracy is specified as 0.65 dB for all off-nadir angles. For a height difference of 500 m, the error considerably exceeds the requirements for the absolute radiometric accuracy, yet (see figure below). An elevation angle dependent correction approach seems to be inevitable.
Best case means that the two back scatterers, which have the same range distance but differing height and horizontal positions, each lie symmetrically on either side of the beam center by half the elevation difference angle. Worst case means that one of the back scatterers lies at the -3 dB margin (i.e. at the rim of the beam cone) and the second back scatterer lies closer to the beam center by the amount of the elevation difference angle. Near and far range refer to off-nadir angles of 18.452 and 40.872 deg, respectively.
Partnerships
German Aerospace Center (DLR / DFD)
Publications
- FREY, O.; MEIER, E.; NUESCH, D.; ROTH, A. [2004]:
Geometric Error Budget Analysis for TerraSAR-X, Proceedings of the EUSAR 2004 Conference, Ulm, Germany - JEHLE, M.; FREY, O.; SMALL, D.; MEIER, E.; NUESCH, D. [2004]:
Improved Knowledge of SAR Geometry through Atmospheric Modelling, Proceedings of the EUSAR 2004 Conference, Ulm, Germany