A Comparative Study of Radar Stereo and Interferometry for DEM Generation

1
A Comparative Study of Radar Stereo and
Interferometry for DEM Generation

• M. Gelautz(1), P. Paillou(2), C. Chen(3), and H.
  Zebker(4)
• (1)Vienna University of Technology, Vienna,
  Austria
• (2)Observatoire Astronomique, Bordeaux, France
• (3)JPL, Pasadena, CA
• (4)Stanford University, CA
• Email gelautz_at_ims.tuwien.ac.at

2
Introduction and Motivation

• Growing availability of interferometric and
  stereoscopic (multi-incidence angle) data sets
  from single and multiple sensors.
• Most existing studies utilize either stereo or
  interferometry exclusively do not provide a
  direct comparison of the two techniques in
  application to one and the same test site.

3
Technique ReviewStereo and Interferometry
Fig1. Imaging geometry.
4
Test Site and Data Set

• Asal Rift, Republic of Djibouti, East Africa.
• Spot reference DEM derived from optical stereo
  analysis and topographic maps.
• ERS-2 interferometric image pair (raw).
• Radarsat stereo image pair
• Radarsat interferometric image pair (SLC).

5
Radarsat Stereo Image Pair
(a) 28 std2
(b) 37 std4
Fig2. SAR stereo pair of the Asal Rift test site
(260 km2), illuminated from the right.
6
Radarsat Interferometric Data Set
(a) Interferogram (after flattening)
(b) Coherence map
Fig3. Radarsat InSAR data acquired during
descending orbits (24 days temporal baseline, 54
m height ambiguity.)
7
ERS-2 Interferometric Data Set

• Magnitude image

(c) Coherence map
(b) Interferogram (after flattening)
Fig4. ERS-2 InSAR data acquired during ascending
orbits (35 days temporal baseline, 17 m height
ambiguity.)
8
Algorithms

• Stereo Matching
• Correlation-based matcher with edge-based
  pre-processing.
• (PAILLOU, P., and GELAUTZ, M., 1999, Relief
  reconstruction from SAR stereo pairs the
  "optimal gradient" matching method. IEEE
  Transactions on Geoscience and Remote Sensing,
  37, 2099-2106.)
• Phase Unwrapping
• Dynamic-cost cycle-canceling (DCC)
• (CHEN, C., and ZEBKER, H., 2000, Network
  approaches to two-dimensional phase unwrapping
  intractability and two new algorithms. Journal of
  the Optical Society of America A, 17, 401-414.)
• DEM Merging
• Averaging algorithm employs a filtered coherence
  map along with user-defined weights.

9
DEM Reconstruction
SPOT reference DEM
Radarsat stereo DEM
Stereo/InSAR merged DEM
ERS-2 interferometric DEM
10
Difference DEMs
InSAR
Merged
Stereo
Radarsat
ERS-2
11
Stereo/InSAR Error HistogramsERS-2 / Radarsat
Fig5. Error histograms of ERS-2 InSAR, Radarsat
stereo, and merged DEM.
12
Scatter PlotsHeight Error vs. Coherence
ERS-2 InSAR
Radarsat Stereo
Merged DEM
13
Results
14
Summary

• The more robust stereo reconstruction was able to
  capture the general terrain shape, but finer
  surface details were lost.
• In regions of high coherence, the interferometric
  DEMs were of much better quality than the stereo
  result.
• Pronounced interferometric phase unwrapping
  errors could be suppressed by combining the
  stereo and InSAR results.

15
Acknowledgements

• This study was performed under the ERS AO3-245
  project.
• The Radarsat data were provided by ADRO-373.