Science Service System

Summary of Proposal MTH0835

TitleOn the use of SAR High Resolution Spotlight data for Persistent Scatterers and Tomographic analysis
Investigator Poncos, Valentin - Kepler Space Inc. / University of Calgary, SAR systems and exploitation
Team Member
Section Froese, Corey - Alberta Geological Survey, Geological Hazards
Associate Professor Collins, Michael - University of Calgary, Geomatics
Geologist Dehls, John - Alberta Geological Survey, Geohazards
SummaryThe objective of the proposal is to combine Radarsat-2 Spotlight mode data with TerraSAR-X Spotlight mode data in oder to better see into the pixel and separate multiple targets contributions. Radar image detection is essentially a 2D process involving the focusing of the backscattered energy into targets located at a specific range, asssuming flat topography. Therefore, targets located at non-zero heights are focused in the wrong place, generating layover effects. Having multiple interferometric images available, it is possible to differentiate between targets located at the same range but distinct heights. This makes possible the 3D focussing of an image and correction of the layover effects. The drawback of the spaceborne SAR systems is that the baseline spread and sampling is not optimal for tomography; TerraSAR-X has a tight orbital tube. Radarsat-2 has a wider orbital tube but it is located at a higher altitude. The acquisition of 20 Spotlight Radarsat-2 images started about 6 months ago. It is proposed simultaneous acquisition of TerraSAR-X data in order to faster create a large dataset with minimum deformation information and a denser baseline sampling in order to improve the beamforming process. The 11 days pass of TerraSAR-X should make possible the acquisition of 25 images by the time the Radarsat-2 dataset was acquired. DLR and IREA suggested that a minimum of 25 images should be used for better Tomographic results. The work is done as part of a research project with U. of Calgary. Kepler Space Inc. will support the funding for this work. U. of Calgary is in the process of applying for new sources of funding.
Detailed reportThis is an excerpt from Valentin Poncos PhD thesis - the full thesis will be available after publication. The expansion of steep areas to their 3-D topography was the object of this research; in these areas the 2-D SAR spatial resolution decreases from meters to thousand of meters squared. In order to correct this resolution loss, back-scatterers mixed in these pixels needed to be separated based on a third dimension, their height. In this way, back-scatterers detected at the same range (time) bin by the SAR system can be correctly localized in space by their true height. The technique of restoring the 3-D properties of an area from 2-D measurements is called Holography. A particular case of Holography when only slices of the continuous 3-D space can be restored is called Tomography (or TomoSAR when applied to SAR system). In the current literature the problem of extracting the 3rd dimension from the data (thickness or height profile) was addressed in various ways, depending on the SAR operating frequency and properties of the illuminated ground (vegetation or infrastructure), both at the SAR system level and (mostly) at the SAR data level. At the SAR system level, knowledge of the actual systems and processing techniques (InSAR, PSInSAR) are required in order to identify and improve components relevant to the TomoSAR technique. At the SAR data level, knowledge of SAR applications and general spectral estimation techniques are required. This research addressed the SAR system level approach, focusing on understanding and improving SAR technology rather than post-processing techniques. In Chapter 2 a theoretical background relevant to this issue was presented in more detail. TomoSAR is a technique based both on amplitude and phase information. The phase property of the SAR varies much faster than the amplitude property and also has more weight in TomoSAR; it is important to be able to follow the eventual phase quality loss as the SAR data is transformed by the hardware receiver system, SAR processor and coherent SAR applications. Chapter 2 enumerated possible phase error sources related to the SAR hardware, propagation media, interaction with the ground and coherent data processing. The phase value of any given pixel contains information from multiple sources. Only one of the phase components is useful for TomoSAR, the rest needed to be removed. Techniques to extract the useful phase component were presented in this chapter. Similar techniques were used later on to extract the equivalent phase noise from real data acquired with Radarsat-2 and TerraSAR-X. Also in Chapter 2 the TomoSAR technique was described; this was used in the implementation of the tomographic aperture synthesis necessary to separate dominant back-scatterers and measure their heights. Chapter 3 described the methods developed to transition the 2-D SAR impulse response to the proposed 3-D version. In this process additional factors related to the location of a back-scatterer in the 3-D space of slant range - azimuth - height were identified and introduced in the new 3-D SAR impulse response expression. After the analysis of the SAR system noise components and of the structure of the TomoSAR design a synthesis of the relevant parameters to the height resolution of SAR data was used to implement the TomoSIM algorithm. This algorithm was used to simulate different SAR system scenarios and feed them to the TomoSAR process in order to observe the behaviour of the height resolution of the SAR data. In the same Chapter 3 a connection between the theoretical simulation and real SAR systems was made through the estimation of an equivalent phase noise figure from data acquired over the test site in Edmonton, Canada. This noise figure was added in the TomoSIM and individual SAR noise components were adjusted in order to accommodate this noise in such a way that the common backscattering scenario illustrated in Figure 6 would be resolved by TomoSAR. Chapter 4 presented results from applying the methods described in Chapter 3. TomoSIM was used to model the SAR system response in the height direction for a number of different scenarios. Due to the large number of parameters involved in simulation of the SAR data in the height direction the testing experiments were extensive. A number of nine experiments using a single dominant back-scatterer and eight experiments using two dominant back-scatterers in a pixel under various SAR system and noise scenarios were presented. Orbits scenarios relevant to the Radarsat-2 and TerraSAR-X systems were used. Carrier frequencies were chosen to match the Radarsat-2 and RCM one side and TerraSAR-X on the other side. Baseline scenarios were chosen to initially match the Radarsat-2, TerraSAR-X and Tandem-X flight paths; afterwards they were modified in order to achieve the desired height resolution permitting the separation of two targets located at a 15-30 [m] height difference between them. Perpendicular baseline spread, sampling and estimation accuracy were analyzed and through multiple experiments the noise associated to these parameters was tuned in such a way to retain the capability of the tomographic aperture to distinguish the two targets. Similarly, the parallel baseline spread, sampling and estimation errors were modeled; it was shown that the parallel baseline spread and sampling are not very important to tomographic accuracy (even large values in the experiments did not affect the aperture resolution) because they induce constant phase offsets for each interferogram that can be corrected through calibration with a common reference point. The parallel baseline estimation error was shown to be very important (in the experiments only millimeter level errors were used - anything beyond that would destroy the aperture). Fortunately these errors are constant for each interferogram and can be also compensated with a very careful phase calibration (pick multiple high SCR references instead of one and calculate a least square solution). The acquisition of a large dataset suitable for TomoSAR requires at least one-two years, time in which the ground can be displaced. The displacement phase can degrade the tomographic aperture. The aperture is synthesized by sorting the repeat-pass data along the perpendicular baseline. After the sorting process the acquisition dates of the scenes become random. A ground displacement process is usually a progressive one (increases in time or it is harmonic); by sorting the data along the perpendicular baseline the displacement process along the tomographic aperture becomes random. Therefore the displacement was modeled as a random process in TomoSIM. From the experiments it was shown that even small displacements (1-2 cm/year) could severely degrade the tomographic aperture. Two solutions are possible: one at the data processing level by removing the deformation information using the PSInSAR technique and another one at SAR system level by increasing the repeat-pass frequency or using multi-static SAR systems. Also in Chapter 4 a bridge between theoretical simulation and real data was crossed by estimation an equivalent phase noise figure from the data. This was done by using the PSInSAR technique on two datasets acquired in Edmonton, one by the TerraSAR-X system and another one by the Radarsat-2 system. The total noise of the detected single dominant back-scatterers in a pixel was measured in the best case scenario, for high SCR back-scatterers close enough to each other to discard atmospheric effects and relative motion. From the total noise an equivalent phase noise figure was estimated by reverse engineering height errors to individual interferograms phase noise (varying as a function of perpendicular baselines). This equivalent phase noise figure (different for TerraSAR and Radarsat-2) was used to create realistic noise models for the two dominant back-scatterers in a pixel scenario. This noise model was fed to the TomoSIM simulator to generate the noisy response of the two targets in a pixel. This noisy response was further introduced in the TomoSAR algorithm and the tomographic aperture was synthesized. The target detector was defined as the half amplitude threshold on the aperture. The conclusion of these experiments was that the total noise of the current Spotlight mode Radarsat-2 and TerraSAR-X data backscattered by hard targets as buildings is too low to critically distort the tomographic aperture. In the end, with the more realistic noise information, a few more experiments were run in order to synthesize for both Radarsat-2 and TerrSAR-X an optimal tomographic aperture that would be able to solve the backscattering scenario in Figure 6 with minimum extra requirements from the SAR system. From these two experiments it was concluded that Radarsat-2 would require a minimum perpendicular baseline spread of 1500 [m] with a deviation from uniform perpendicular baseline sampling of maximum +/-50 [m] and perpendicular baseline estimation accuracy within a few centimeters. For TerraSAR-X the baseline spread criteria can be relaxed; only a 500 [m] baseline would suffice to reach the same tomographic capabilities as Radarsat-2. Any high-accuracy measurement technique based on remote sensing techniques (acoustic or electromagnetic) needs to pass the following number of key stages in order to advance and reach maturity: 1. become coherent (control the phase of the wave) for 2-D sub-wavelength measurements (sonar, radar, laser, seismic). 2. become interferometric - to use phase fringes to extract both range and height of a single target (laser Interferometry, InSAS and InSAR) 3. become tomographic - to use 3-D phase fringes to extract slices (height profiles) of the illuminated media (CAT Scan, seismic tomography, TomoSAR). 4. become holographic - to use 3-D phase fringes to extract the full 3-D volume (laser, electromagnetic holography). The current space-borne SAR systems were developed mostly as imaging systems (no ranging properties are used) or interferometric systems (stage 2). This research analyzed current SAR systems from the perspective of understanding the steps required to advance space-borne SAR technology from level 2 to 3. The importance of level 3 in SAR is underlined by the recent approval of the ESA BioMASS concept for the Earth Explorer 7 mission scheduled to be launched in 2020. At the core of this mission TomoSAR applications will create tomograms of the forests for the canopy height and density measurement, necessary for biomass calculation. More advanced work using SAR airborne systems advance from stage 3 to 4 is pursued at DLR (German Space Center). At IGARSS 2013 the first paper about SAR Holography was presented, being also the winner of the student contest. SAR Tomography will continue to be developed to the point when it will become operational. The last step, SAR Holography, will signal the entrance of the SAR system into the club of highest accuracy measurement systems such as electronic microscopy.

Back to list of proposals

DLR 2004-2016