Jürgen Holzner

Traffic monitoring using SRTM along-track interferometry

In February 2000, the shuttle radar topography mission SRTM was flown as the first single-pass SAR interferometer in space. The goal was to obtain a global digital elevation model. Therefore, a secondary receive-only antenna on the top of a 60 m long deployable mast supplemented the primary transmit and receive antenna mounted in the shuttle cargo bay. Due to mechanical constraints, the interferometric baseline did not only consist of the terrain height sensitive across track component but also contained an along track component of 7 m. The resulting time lag between the data acquisitions by the two antennas causes phase differences, which are proportional to the line-of-sight velocity of moving targets. Since this system ability has been expected prior to the SRTM launch, a traffic monitoring experiment has been carried out during the mission, which showed a surprisingly good agreement between the GPS velocity measurements on ground and the measurement from space. A strategy to exploit the along track interferometric (ATI) phase and the coherence together with the azimuth displacement of moving vehicles in focused SAR images has been developed and applied on sample test sides for traffic monitoring on highways.

SRTM X-SAR calibration results

The paper presents the results of the Shuttle Radar Topography Mission (SRTM) X-band calibration at DLR as of April 2001, the time of writing. While some areas may be subject to further improvement, the overall image will not change significantly. We summarize the various calibration issues, the methods applied and the results obtained. Addressed topics are: Timing calibration, SNR and coherence, motion analysis and instrument phase errors. The qualitative and quantitative effects of the distortions above on the DEM quality are discussed.

RADARSAT ScanSAR interferometry

ScanSAR interferometry is an efficient technique for topographic mapping or surface change monitoring of large areas. The feasibility of ScanSAR interferometry has been demonstrated in theory and simulations before. The authors show the first ScanSAR interferogram from real RADARSAT data. In a first example, an interferogram derived from two ScanSAR data sets is presented. In a second example, an interferogram is formed from a ScanSAR a one strip-map data set. In this case, the azimuth synchronization problem involved in the acquisition of ScanSAR interferometric pairs is circumvented. Different interferometric ScanSAR algorithms are discussed.

Interferometric DEMs in alpine terrain-limits and options for ERS and SRTM

With ERS tandem data being todays primary interferometric data source, DEM generation is currently limited to moderate terrain. In general, layover and shadow affect the visibility and resolution of SAR sensors in rugged terrain. As a consequence, classification and phase unwrapping of such data is difficult. Shuttle Topography Mission SRTM X-SAR data is expected to give better results due to its viewing geometry and the high coherence, but it might still suffer from the problem under extreme conditions. The authors perform a quantitative analysis of the radar illumination conditions, using a digital elevation model. The potential benefit of combining data from different viewing geometries like incidence angle and orbit inclination are shown.

Phase unwrapping of low coherence differential interferograms

The authors are presenting a new method to unwrap the phase from irregularly sampled differential SAR interferograms. Often such interferograms are used to observe land surface deformation in the order of centimeters over long time intervals, say up to years. However, in vegetated regions, the surface tends to decorrelate within this time frame and complicates a quantitative long term analysis. This is especially true for the short C-band wavelength of the space borne sensor ERS, the primary data source for radar interferometry. The authors are presenting a variation of Costantinis minimum cost flow method, that avoids low coherence areas and uses isolated man made or natural point scattering objects instead in order to derive the deformation parameters of large scale land surface changes. The method implies automatic selection of stable targets, analysis of their quality, unwrapping of the target phases. The paper focuses on the phase unwrapping method, an adapted version of the minimum cost flow approach. Results are presented both from simulation and from a sample data set over a land subsidence area.

Filtering of interferometric SRTM X-SAR data

In February 2000, the Shuttle Radar Topography Mission acquired a global digital elevation model (DEM) within only 11 days. During the years 2000 and 2001, extensive testing and calibration activities followed at DLR. One of the questions during that phase was, which kind of interferogram filtering should be performed. We analyzed different types of filters from the perspective of an operational DEM processing chain. Our paper describes the filters investigated and the arguments that finally led us to the decision to implement a relatively simple Gaussian smoothing in the time domain. This filter performed best in many disciplines and produced the smallest artifacts.

A linear Kalman filter approach for estimation of a vehicle's motion parameters using range-Doppler tracking and road information

For the processing of ISAR images of curving vehicles, the accurate knowledge of the target position as well as the aspect angle rate is essential. Especially when the targets are non-cooperative, there are no measurements like GPS positions available. This paper presents a Kalman filter based algorithm which fuses radar measurements of distance and velocity in line of sight with road information. The result is a set of parameters containing target position, velocity and acceleration, each in Cartesian coordinates. The aspect angle can then be obtained by calculating the angle between target velocity vector and the slant range vector.

ScanSAR interferometry for RADARSAT-2 and RADARSAT-3

Among the interferometric modes and applications of RADARSAT-2 and RADARSAT-3, ScanSAR interferometry is interesting because of its wide swath imaging capabilities. It allows efficient mapping of the earths surface topography and its changes, and poses several interesting requirements on instrumentation, data acquisition, and processing. These are discussed for RADARSAT-2 and RADARSAT-3. The signal properties of ScanSAR data are derived, and an interferometric ScanSAR product is proposed. A digital elevation model, derived from RADARSAT-1 ScanSAR data, is presented and demonstrates the feasibility of ScanSAR interferometry for topographic reconstruction.

Geometric performance of ENVISAT ASAR products

We describe validation measurements of the geometric accuracy of ASAR images, measured redundantly via independent methods. Our tests include image (IM), alternating polarization (AP), and wide swath (WS) mode acquisitions over a variety of test sites. ASARs slant range products (IMS/APS) require a slightly different validation methodology than ground range precision (IMP, APP) and medium resolution products (IMM, APM, WSM). A third approach is required for ellipsoid-geocoded products (IMG, APG). The most highly accurate validation is possible with single look complex (SLC) data (IMS and APS products), as all other product types lose resolution during multilooking. For a library of ground control points (GCPs) including map features such as bridges or road intersections, as well as (where available) transponders and corner reflectors, we use surveyed or map-measured position information (together with the delay value in the case of transponders) to solve the zero-Doppler iteration and predict the position of the GCP as an azimuth and slant range coordinate in the radar image. In the case of ground range products (e.g. IMP, APP, IMM, APM, WSM) the predicted slant range value is additionally transformed by a slant to ground range transformation tro determine the predicted image coordinate. The GCP feature is then either measured by inspection of a detected image, or localized automatically within the neighborhood of the prediction. GCPs are measured within the radar geometry image products, derivative geocoded products, and topographic maps, providing their measured map, radar geometry, and nominally geocoded GTC locations. Radar image locations are compared to map reference values and statistics of differences are tabulated. We compare the accuracies of the estimates achievable using transponders and map GCPs. Based on the suite of products (and accompanying orbit information) available to us, we establish a methodology for estimating a preliminary sampling window start time bias. The multiple validation and estimation techniques used ensure robust determination of ASAR geolocation accuracy.

Analysis of interferometric signals based on coherence and power spectral density

It the paper of Holzner and Bamler (2002) it was proven that two interferometric burst-processing options are equivalent. In order to address this question interferogram sample coherence and power spectral density were applied. This paper proposes to use this two expectation values as a general instrument to explore various SAR modes and their interferometric signal properties. Moreover, this method is a way to represent and investigate the behavior of the interferometric signals in a unified and concise way. In turn, this approach will help to design new algorithms for SAR interferometry. Burst-mode interferometry serves as an example. All the known properties of interferometric burst-mode signals - for the cases of non-zero fringe frequency, burst pattern misalignment, and misregistration - can be inferred very easily. In order to amplify the spectral behavior of interferometric burst-mode signals the results are contrasted/referenced to the ones obtained for strip-map interferometry. On this basis, the expectations for spotlight interferometry can be predicted and discussed.