In: Stilla U, Rottensteiner F, Paparoditis N (Eds) CMRT09. IAPRS, Vol. XXXVIII, Part 3/W4 — Paris, France, 3-4 September, 2009
THEORETICAL ANALYSIS OF BUILDING HEIGHT ESTIMATION USING SPACE
BORNE SAR-INTERFEROMETRY FOR RAPID MAPPING APPLICATIONS
Stefan Hinz 1 , Sarah Abelen 2
'Institute of Photogrammetry and Remote Sensing, Univeristät Karlsruhe, Kaiserstr. 12, 76 128 Karlsruhe
2 Remote Sensing Technology, Technische Universität München, Arcisstr. 21,81 333 München
Stefan.Hinz@ipf.uni-karlsruhe.de
Commission III, WG 5
KEY WORDS: Space-borne Interferometrie SAR, Building Heights, TanDEM-X. CosmoSkymed
ABSTRACT:
The great potential of space-borne SAR images for semi- or fully-automatic mapping of topographic features has been shown by
many approaches. While most of them focus on 2D mapping of topographic features, some preliminary research on the complex task
of automatic delineation of 3D information in urban environments has been initiated in recent years. In this paper, we analyze the
capabilities of new space-borne interferometric SAR missions - in particular the German TanDEM-X mission - with respect to their
potential of deriving building heights. To this end, we summarize the mathematical framework and carry out a thorough analytical
accuracy analysis involving various sensor and scene parameters.
1. INTRODUCTION
The new class of space-borne high resolution SAR sensors such
as TerraSAR-X, SAR-Lupe or Cosmo-SkyMed is able to
provide SAR images of l-3m spatial resolution or even below
in special spotlight modes. Naturally, the development of
methods to automatically derive detailed cartographic
information from this kind of data is a major issue driven by
these missions. Since SAR is largely independent from
illumination and weather conditions, it is furthermore an
attractive imaging technology for acquiring area-wide
information of regions hit by disasters such as floodings,
landslides, or earthquakes.
The great potential of space-borne SAR images for semi- or
fully-automatic mapping of 2D topographic features has been
shown by many encouraging approaches, e.g., (Negri et al.,
2006; Frey & Butenuth, 2009) for delineation of roads and
(Jäger et al. 2007; Hänsch & Hellwich, 2008) for classification
of agricultural features, just to name few recent ones. The
derivation of 3D features is however more difficult, since these
current civilian space-borne systems have only limited
interferometric capabilities. While the acquisition of along-track
interferometric image pairs is possible by programming special
RADAR imaging modes (e.g. DRA mode or Aperture
Switching mode for TerraSAR-X (Runge et ah, 2006)) enabling
the detection of moving objects (Suchandt et ah, 2008; Weihing
et ah, 2008), none of the current civilian space systems is
equipped with an across-track interferometer, which would
provide the basis for deriving topographic heights (Bamler &
Hartl, 1998; Cumming & Wong, 2005). The necessary across-
track baseline is only given when forming an interferogram of
two SAR acquisitions taken from the same orbit yet at different
passes of the satellite, and thereby relying on the positional
variation of the orbits. It is clear that the resulting
interferograms suffer from decorrelation depending on temporal
variability of the objects under investigation.
This situation will change once TerraSAR-X will be
accompaigned by a second, quasi-identical SAR satellite in late
2009, leading to the TanDEM-X mission. Both satellites will fly
almost in parallel forming a helix-like orbit pair (Zink et ah
2006). This configuration allows to acquiring SAR image pairs
with variable across-track geometry resulting in a significantly
improved interferometric coherence. The great benefit of single
pass across-track SAR interferometers has been intensively
studied in the context of the Shuttle Radar Topography Mission
(SRTM). Despite of the limited spatial resolution of SRTM data
(approx. 25m), it was possible to compute a global digital
elevation model with standardized height accuracy of few
meters, see, e.g., the comprehensive overview given in (Rabus
et ah, 2003).
TanDEM-X will deliver high coherence interferometric data of
the meter class. Although the mission is mainly designed to
generate accurate digital elevation models satisfying HRTI-3
standards (Zink et ah, 2006), it can be expected that this kind of
data opens up a much wider field for specialized methods for
3D mapping of topographic features. The automated derivation
of building heights or even the detailed reconstruction of
buildings is certainly an important application amongst these.
Apart from the improved spatial resolution, a major difference
between TanDEM-X and SRTM is the variable across-track
baseline of TanDEM-X, whereas the baseline of SRTM was
held quasi-constant due to the second antenna mounted at a
60m boom (and neglecting periodic baseline variations as
consequence of thrusting). Hence, a thorough analysis of
accuracy aspects of height estimation under the given flexibility
of TanDEM-X is a key issue.
Following questions should be answered by the analysis:
Which accuracy level in terms of building height
estimation can be reached with interferometric data as
it will be provided by TanDEM-X?
Is this accuracy sufficient to derive object specific
information for rapid mapping in the context of crisis
management? Such information may comprise, e.g.,
o the number of floors to estimate the amount
of people living in a house
o attached building parts of different height
o the roof type (flat roof, saddle roof, etc.)
How would the accuracy improve, if external data
from GIS is included (e.g. digital building footprints)?