SAR Interferometric DEM quality assessment
Aérospatiale Espace & Défence, Cannes (France)
Ecole Supérieure des Géomètres et Topographes,
Institut Géographique National, Saint-Mandé (France)
aptitude of SAR interferometric acquisition for relief mapping has been
demonstrated, and illustrated on a few test sites. Unfortunately, the validation
of this mapping technique is still very limited, for three main reasons
- First, the quality assessment of each interferometric DEM is generally
performed using questionable data sets, i.e. either a small number of ground
control points with limited statistical representativity, or a so-called
reference DEM (obtained from a SPOT stereo pair or from an official data
base) which is often less accurate than the interferometric DEM itself.
- Second, very few experiments have been carried out completely, i.e.
including height computation and DEM geocoding. Therefore, it is difficult
to predict the performances of SAR interferometry in general terms without
taking into account the influence of a particular landscape (vegetation,
atmosphere, slope...) or particular acquisition conditions (sensor parameters,
- Third, DEM quality assessment is often limited to a single criterion,
namely, elevation standard deviation, which is very easy to evaluate. This
criterion is very relevant for orthorectification but not for most DEM
applications, in particular geoscientific applications. Therefore, user
needs have to be expressed more clearly before a user-oriented quality
assessment can be undertaken.
- These limitations will be analysed in our presentation. We will
propose solutions for overcoming them, based on a simulation-based concept
already published (Polidori & Armand 1995)
and we will present some preliminary results.
- Keywords: SAR,
interferometry, DEM, quality, errors
To overcome this problem, we have developed a validation environment
dedicated to testing new algorithms or new SAR systems and to quantitatively
evaluate the aptitude of existing interferometric processing chains to
map relevant topographic parameters (altitude but also slope or terrain
motion). The approach we use is based on artifact modelling and image simulation.
Indeed, studies performed by Aerospatiale and ESGT have shown that SAR
image simulation is a powerful tool for the validation of relief mapping
techniques such as SAR interferometry, stereo-radargrammetry or shape-
from-shading (Polidori & Armand 1995).
We have also successfully used this approach to validate other cartographic
applications, e.g. change detection in radar images, building extraction
from optical stereo images or spectral sensing from future MERIS data.
Simulation-based validation of radar interferometric processing
(illustrated on Figure 1) has three main advantages
: * it allows a great variety of situations, i.e. a variety of both SAR
systems and landscapes, so that a new algorithm can be tested in many cases
and not only on a particular data set ; * it can be handled in a parametric
way so that the impact of any parameter on the resulting accuracy can be
evaluated (by parameter we refer to both imaging parameters related to
sensor, SAR processing or acquisition conditions - and landscape parameters
such as slope, roughness, moisture but also atmospheric refraction index)
; * it relies on an input landscape which can be considered as an exact
and dense ground truth, required to derive error maps and therefore to
analyse the relationship between the behaviour of an algorithm and local
1 : Simulation-based validation environment for radar interferometric processing.
In the case of SAR interferometry, the simulation procedure must be
designed in such a way that all phase effects can be taken into account,
namely : - orbit ; - platform stability ; - clock ; - atmosphere and ionosphere
; - raw signal compression and decompression ; - SAR processing ; - terrain
elevation ; - terrain slope and curvature ; - roughness (i.e. spatial organization
of scatterers) ; - volumetric scattering ; - subsurface penetration ; -
temporal variation of ground parameters.
Since it would be very time consuming to consider all these effects
in a rigorous raw signal simulator, we have implemented a simplified simulator,
fully dedicated to interferometry, in which the SAR impulse response is
modelled in a separate simulator.
Some results are presented in fig.2A reference
map is presented in fig.2a, based on a digital
elevation model and a synthetic, manually drawn land use map. Sensor parameters
corresponding to the ERS case have been used. Fig.2b
represents the raw interferogram obtained with a 64 meter baseline. Phase
unwrapping, elevation computation and finally geocoding have led to the
output map presented in fig.2c. A visual comparison
of the input map and the output map is not very relevant since the data
are very similar. Reversely, computing elevation differences leads to a
very interesting error map (fig.2d), on which
the local behaviour of our algorithms can be revealed. Error histograms
can be computed, even over restricted areas, which would not be possible
when validating with few unaccurate ground control points.
Fig.3 shows similar results obtained with
a greater baseline (138 m). The effect of the baseline can be evaluated
not only in average but also locally.
: input map
: raw interferogram
: output map
: error map
Fig. 2 : Results obtained for baseline = 64 m
: raw interferogram
: error map
Fig. 3 : Results obtained with baseline = 138 m
In conclusion, a validation tool based on a dedicated image simulator
is available. We use it to evaluate and improve our own interferometric
processing chain, but we can propose it to other research groups who wish
to evaluate the performances of their algorithms or test them in very special
conditions, for instance for sensors that have not been launched yet. In
a very near future, we are planning to use the same approach to validate
slope mapping algorithms in the frame of the ERS tandem mission.
- Polidori L. & Armand P. 1995:
- On the use of SAR image simulation for the validation of topographic
mapping techniques. EARSEL, Advances in Remote Sensing, Vol 4,
No 2, pp. 40-48.
Keywords: ESA European
Space Agency - Agence spatiale europeenne,
observation de la terre, earth observation,
satellite remote sensing,
teledetection, geophysique, altimetrie, radar,
chimique atmospherique, geophysics, altimetry, radar,