From Radar to Reality.Associating persistent scatterers to corresponding objects

Research output: ThesisDissertation (TU Delft)

32 Downloads (Pure)

Abstract

Multi-epoch Synthetic Aperture Radar Interferometry (InSAR) is widely used to estimate displacements of selected scatterers from phase observations. However, their interpretation needs a connection to objects in the real world.

To associate InSAR scatterers to their corresponding geo-objects, it is necessary to (i) accurately estimate the phase center of radar scatterers in radar coordinates, (ii) precisely position the scatterers in 3D geographic coordinates, and (iii) satisfy the constraint that these positions need to be physically realistic. This study addresses these three requirements.


The effective phase center of a scatterer is not situated at the nominal position of the pixel. As a result, scatterers are evaluated at the wrong position and the reference phase calculated at that location will be biased. We evaluate the influence of this sub-pixel position on the geolocation of the scatterer and its deformation quality for various satellite platforms. A method to locate the phase center of the dominant scatterer is developed and is applied to a stack of TerraSAR-X, Radarsat-2, and Sentinel-1 images. The sub-pixel correction shows to be significant for improving the geolocation, up to a few meters—especially for planar (horizontal) precision. It is only of limited influence for the displacement estimation and more relevant in the case of large orbital baselines.

Even after sub-pixel correction, the position of scatterers in an earth-centered, earth-fixed geodetic datum is often in order of a few meters, which is not always sufficient to physically link the scatterer to a geo-object. We evaluate four approaches for correcting this positioning bias,i.e., (i) an advanced geophysical correction, (ii) the single-epoch deployment of a corner reflector (CR), (iii) a multi-epoch deployment of a CR, and (iv) a correction using a high-precision digital surface model (DSM). The positioning performance of these approaches is analyzed from the aspects of practicability, reliability, and precision with TerraSAR-X and Sentinel-1 data. We show that while the multi-epoch CR approach achieves the best positioning results, the DSM-assisted correction is able to obtain comparable results if a high precision DSM is available, better than DTED-4.


The position of the estimated geometric phase center may differ from the position of the physical phase center. We use ray-tracing to predict the position of point scatterers using generic 3D models, and match them with the detected point scatterers from a stack of TerraSAR-X images. We find that the majority of detected scatterers appears to be positioned at their correct physical location. Moreover, many point scatterers correspond to multiple scattering mechanisms—more than half of the identified scatterers correspond to double- or triple-bounce scatterers. The mismatch between the geometrically estimated position and the signal source occurs mainly for multiple scattering: fourfold and more. This shows that the bounce levels of the scatterers are a relevant attribute to understand and interpret the displacements of persistent scatterers.

In general, we conclude that sub-pixel correction and positioning bias correction should be included in default InSAR data processing, and that the majority of detected scatterers are positioned at physically realistic locations.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Delft University of Technology
Supervisors/Advisors
  • Hanssen, R.F., Supervisor
  • Lopez Dekker, F.J., Advisor
Award date2 Sep 2020
Print ISBNs978-94-6384-128-3
DOIs
Publication statusPublished - 2020

Keywords

  • satellite radar interferometry
  • time series InSAR technique
  • sub-pixel correction
  • precise point positioning
  • corner reflectors
  • digital surface model
  • multiple scattering
  • ray-tracing

Fingerprint Dive into the research topics of 'From Radar to Reality.Associating persistent scatterers to corresponding objects'. Together they form a unique fingerprint.

Cite this