International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004
2. TEST AREAS AND GROUND REFERENCE DATA
The test area chosen by DLR is a region of about 40 x 50 km? in
the southeastern part of Bavaria. The elevations range from 400
to 2000 meters in a mostly hilly, post-glacial landscape
including some lakes and also mountains of the German Alps.
This selection allows the comparison of DEM for different land
surface shapes, including forest and steep terrain.
The ground reference data selected for this test area are the
following (see also fig.1):
e Four regions have a grid spacing of 5 meters and an overall
size of about 5 km x 5 km, derived from airborne laser
scanning. The height accuracy is better than 0.5 meter.
e One region (area of Inzell, total: 10 km x 10 km, 25 meter
spacing) consists partly of laser scanner data (northern
part). The height accuracy is better than 0.5 meter.
The southern part of the DEM is derived from contour-
lines 1:10 000. The height accuracy is about 5 meter due to
the mountainous area.
eo A large region (50 km x 30 km) is covered by a coarser
DEM with 50 meter grid spacing and height accuracy of
about 2 meters, derived from aerial photogrammetry
e The exact locations of 81 GCP (fix points) are listed in a
pdf document.
Bavaria Nonh HRS t «tho
DEM-01
Figure 1. Location of the test area at the most southeastern
part of Germany, location of the SPOT-HRS scenes with
DEM reference sites and ground control points.
3. PREPROCESSING OF THE ANCILLARY DATA
The delivered SPOT 5 HRS Level 1A product consists of the
image data in standard TIF format and the meta data in DIMAP
format. The following information is extracted for each CCD
array from the XML ancillary file for further processing:
e satellite ephemeris data containing position and velocity
measured by the DORIS system every 30 seconds with
respect to the ITRF90 (International Terrestrial Reference
Frame 1990) system during the data take and at least four
times before and after image data acquisition,
e corrected attitude data with respect to the local orbital
coordinate frame measured by gyros and the star tracker
unit ULS with 8Hz, the data are already corrected for
different effects (Bouillon et al. 2003)
e look direction table for the 12000 CCD elements expressed
within the sensor coordinate frame
e data used for time synchronization like line sampling
period and scene center time.
According to the ,,SPOT Satellite Geometry Handbook* (SPOT
IMAGE 2002) Lagrange interpolation of the ephemeris data
and linear interpolation of the attitude data are recommended to
calculate the exterior orientation for each scan line. After
transformation to a local topocentric system (LTS) with a
fundamental point located at the center of the image scene, this
serves as input for DLR’s processing software. For orthoimage
production the exterior orientation is expressed in the Earth
Centered Earth Fixed (ECEF) WGS84 Cartesian frame.
4. IMAGE MATCHING
Matching of the two images is performed purely in image space
with DLR software. Details on this software are described in
Lehner et al. 1992. It relies on a 7-step image resolution
pyramid and applies intensity matching in two forms:
normalized correlation coefficient for pixel accuracy and
subsequent local least squares matching (LLSQM) for
refinement to sub-pixel accuracy (for mass points 0.1 to 0.3
pixel standard deviation). Interest points are generated with the
Forstner operator and the homologous points are searched for in
the other image. Only points with high correlation and quality
figure are selected as tie points for bundle adjustment (see
chapter 7) and a less stringent criterion is valid for the usage as
seed points for the subsequent Otto-Chau region growing
procedure for dense matching (Heipke et al 1996). This local
least squares matching starts with template matrixes of 13 x 13
pixels around the seed points with a step of 1 to 3 pixel in each
direction. For cross checking a backward match is performed
for all points found. Some details are described in Müller et al.
2004.
5. ORTHOIMAGE GENERATION AND ACCURACY
ANALYSIS
To get an impression of the absolute and relative accuracy of
the position and attitude data, and to get an estimation of the
necessity to improve the ancillary data by bundle adjustment or
other methods, orthoimages are derived using an already
available DEM.
The inputs for the orthoimage production are the interior
orientation (CCD look angles) the six parameters of the
exterior orientation for each image line (interpolated from the
measured sampling points) and the DEM. In the case of Bavaria
the DEM has been derived by DLR from several ERS 1/2
Tandem pairs, the accuracy is in the order of 5 to 10 meter in
flat and hilly terrain and 10 to 50 meter in mountainous terrain
(Roth et al. 1998). Therefore the more reliable part of the
orthoimages is found north of the foot of the Alps.
The principle of the orthoimage production is based on the
intersection of the actual sensor viewing direction (pointing
vector) with the DEM applying the rigorous collinearity
equation. The orthoimage processor calculates the object space
coordinates of the points within the local topocentric system
and then transforms them to the desired map projection of the
output image using geodetic datum transformation parameters
(Müller et al. 2003). Bilinear resampling to a 10 x 10 m grid has
been performed.
After generation of the two orthoimages without any ground
control information, a check of the accuracy using 20 of the
ground control points has been performed. For the quality
assessment the measurements have been carried out in
bilinearly enlarged orthoimages to achieve sub-pixel accuracy.
Table 1 shows the deviation in x and y direction for the
orthoimages in comparison to the control points.
Inter.
——
diffe
xl, y
x2. y
x3. y
|
9
The
abso!
than
expe
giver
accu
pape
6.
Havi
the e
objec
inter:
inter:
point
regul
The
algor
surfa
terrai
in for
In th
testir
50 kr
the |
HRS
