The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
145
4.3 Planned Test Program and Current Status
The intent is to acquire data in several areas representing
different forest conditions and topography. In all cases,
ancillary data are required as part of the validation exercise.
Selection of the test sites is based upon the availability and
quality of the following data sets:
• Lidar ground data for validation of the PolInSAR-
derived bare-earth DEM,
• X-band data from previous STAR-3/ or TopoSAR
campaigns will be used for corroboration of tree
height,
• Reports on species, tree heights, stem density, DBH,
etc as available from public or private sources
Preliminary tests have occurred, enabling engineering issues to
be addressed and mitigated. SNR at the 25db level has been
confirmed for VV and HH in bare areas. Polarimetric
calibration has led to cross-talk less than -25 db which is
deemed adequate. Polarization imbalances have been corrected
to the 5%/5 degree level (amplitude/phase).
4.4 Preliminary results
Some preliminary results have been obtained from one of the
datasets acquired recently. The test site is in a heavily forested
region containing a pattern of alternate forest and clearcut areas.
The forest in this area is mainly composed of pine trees with
height ranging from 10-25m and with relatively homogeneous
growth in the forested patches. Figure 7 shows a ground photo
looking towards the forest. The forest in this location is quite
dense, with estimated tree heights about 20m on average. A
bare-earth DEM was available for truth (30 cm accuracy quoted
by Terrapoint, the supplier). Also, an X-Band DSM from the
TopoSAR system flown in 2006 (prior to the L-Band
configuration change) was available to provide an under
estimated reference of the canopy height.
process within the forest patches however, reflects the ground
elevation over most of the test area.
In Figure 8 we show the result of merging the two results such
as to preserve elevation continuity across the forest boundary.
The blue line shows the merged elevation profile representing
the terrain while the green line and the red lines are the
approximate canopy and bare-earth elevations from X-Band and
lidar respectively. The merged L-Band DEM remains within 5
meters of the ‘truth’ within the 15-20 meter forest. However it
fails in the sloped region where canopy heights are reduced.
This is possibly because the model is not optimized for trees in
this height range. This issue is to be addressed.
1140
1135
1130
1125
1120
1115
Si
fc mo
I 1105
1100
1095
1090
1085
1080
Cross Section
Figure 7. Ground photo during L-Band data acquisition
■
The coherence (magnitude) optimization and coherence region
algorithms have been applied to the calibrated dataset. The
elevation derived from the optimized coherence magnitude
follows the terrain quite well on the bare/new growth areas but
inside the forest canopy appears to be similar to the X-Band
DSM height. The elevation derived from the phase optimization
Figure 8. Top: Elevation cross-profile of lidar bald-earth (Red),
X-Band DSM (data acquired in April 2006, Green), L-Band
ground elevation (blue). Bottom: Location of the profile line is
overlaid on a color airphoto.
5. SUMMARY AND CONCLUSIONS
The NEXTMap" program, based upon airborne X-Band HH
InSAR, is in the process of creating a 3-dimensional,
homogeneous, seamless database including DSM, DTM and
ORI products, for eighteen countries of Western Europe and for
the USA (excluding Alaska). The DSM is specified at 1 m
RMSE vertical accuracy for 5 m sample spacing, while the ORI
is specified at 1.25m resolution with better than 2 m RMSE
horizontal accuracy. For many applications the combination of
detail provided, national and super-national availability and
shared pricing through license arrangements should produce an
attractive user opportunity. The creation of the product datasets
for both areas is well underway and scheduled for completion,
in the case of Western Europe, for late 2008 and about 1 year
later for the USA. With respect to many market applications
such as visualization, flood risk, and auto safety these products
should occupy a solid niche, relative to alternative technologies.