al. 1982; Müller 1991/. The exterior orientation parame-
ters are determined only for so-called orientation ima-
ges, which are introduced at certain time intervals. In
between, the parameters of an arbitrary image line are
computed from the values of the neighbouring orienta-
tion images, using a Lagrange polynomial (LP). One
advantage of the LP approach is that the coefficients are
the (unknown) exterior orientation parameters of the
orientation images themselves /Ebner et al. 1992/.
In both practical examples four orientation images at a
distance of about 2000 image lines were introduced for
each orbit. The temporal variation was modeled by a 3rd
order LP. XYZ object coordinates for the projection
centres of these orientation images were derived from
the SPOT header information. It was assumed that these
coordinates have. high relative and moderate absolute
accuracy, i.e. that the shape of the orbit is well repre-
sented in these data, but not its absolute position in
space. In order to model these accuracy properties,
observation equations for the coordinates of the projec-
tion centres were formulated including additional offset
parameters. The relative accuracy is expressed by the
standard deviations assigned to the observation equa-
tions; the absolute accuracy is represented by the stand-
ard deviations of the offset parameters.
In the adjustment the ground coordinates of the object
points, the exterior orientation parameters of the orien-
tation images, and the offset parameters were conside-
red as unknowns. They were estimated from the image
coordinates, position observations derived from the
SPOT header data, and coordinates of the GCP. Attitu-
de observations of the exterior orientation parameters
were not considered in the adjustment.
4. PRACTICAL PROJECTS
In this chapter, the results of two projects termed "Hei-
delberg" and "Priorat" respectively are reported in order
to demonstrate the capability of the described approach.
4.1. Heidelber
4.1.1. Project description. The images depict an area
near Heidelberg, Germany (see figure 2). They were
recorded on November 8 and 17, 1988 respectively. The
base-to-height ratio amounts to 0.4. In the image from
November 17 a few clouds in the north west and south
west are visible, the other image is entirely free of clouds.
The Rhine valley can be seen in the west of the scene,
the upper north shows the Odenwald, a mountainous
region with heights of up to 600 m.
834 GCP derived from high altitude aerotriangulation
and well distributed over the images were given. These
GCP have an accuracy of about 3 m in planimetry and 5
m in height. Since for the measurement of image coor-
dinates no adequate digital device was available, these
were measured stereoscopically on an analytical plotter
using film hardcopies provided by SPOT IMAGE. The
large number of GCP provides an excellent means for
independent control.
For image matching ten points out of the available GCP
were selected as starting points. The size of the template
matrix was set to 19 * 19 pixels. From experience the
following thresholds were introduced:
- 0.6 for thé minimum correlation coefficient pmin,
- 0.1 pixel for the maximum semi-major axis of the
error ellipse,
468
P^ Dal pet pump. u ~~
pm — PM - m
