ifr
all
re
le
l=
le
1000
A
Resurs-01 SK
100
= IRS-1Ce
E
= MOMS:028
£
€ (A)
>
E
i
a Space Imaging A
10-9—7 S equivalent to
scan line of
QuickBird 6000 CCDs
OrbView
1^1
1 10 100 1000
Pixel size [m]
Figure 1: Ground pixel size versus swath width of (pre-)
operational and proposed sensors (logarithmic scale)
These numbers can be illustrated even better with the nec-
essary number of scenes for creating a map sheet at a
certain scale and extent. Assuming a target scale of 1 : 25
000 and a sheet of 1m * 1m one or two HR-MOMS-02
scenes are needed, whereas 25 scenes of QuickBird or 49
scenes of OrbView have to be taken into account which
lead to the conclusion that these proposed 1-m-US-sys-
tems are not cost-effective at those scales.
The temporal coverage describes the desired point as well
as the frequency of data acquisition. Again, due to its
experimental or pre-operational character MOMS-02 pos-
sesses limited features. MOMS-02-02/P is planned to stay
in space for 18 months whereby swaths next to each other
have a track difference of three days. The recording time
will be reduced because it has to be shared with other sen-
sors onboard the MIR-station.
Summarizing this, it can be stated that the MOMS-02-
instrument itself offers a good coverage for medium scale
mapping purposes especially in terms of the spatial extent.
The overall mission parameters have yet to be adopted for
a full operational mission which is envisaged for the year
2000 (MOMS-03 mission).
3 GEOMETRICAL ASPECTS
Mapping at certain scales demands for appropriate geo-
metrical accuracies of the data source. After describing
these demands (section 3.1) the strategy for a practical
evaluation will be outlined (section 3.2) and results will be
presented considering the impact of various input parame-
ters (section 3.3).
743
3.1 Geometrical demands
There are no world-wide accepted rules for horizontal and
vertical accuracies for topographical databases so that the
following values only reflect an average. According to that,
for the horizontal accuracy a value of 0.3 mm times the tar-
get map scale should be achieved, e.g. 7.5 m at a scale of
1:25 000. The measure for the vertical accuracy amounts
to 0.3 times the contour interval which itself depends on
the terrain and the specific application. Assuming typical
intervals in the range of 10 to 20 m the accuracy should be
in the order of 3 to 6 m.
3.2 Methodology
The general strategy for a practical evaluation of the geo-
metrical accuracy is to compare coordinates of distinct
control points as derived from MOMS-02 imagery with
given reference coordinates.
Therefore, control points are measured digitally within the
MOMS-02 images in monoscopic or stereoscopic mode.
Stereoscopic measurements are performed by means of a
digital photogrammetrical workstation using the Crystal
Eyes principle (Siebe et.al., 1992). For this, the oblique
channels FW and BW can be used as well as a combina-
tion of the high resolution channel (HR) and one of the
oblique channels which have have to be brought to the
same scale.
Reference coordinates of the control points have been
obtained by digitizing from maps, ranging from 1 : 5 000
(test site Harare, Zimbabawe) over 1 : 10 000 (orthoimage
maps of Dubai-City, United Arab Emirates) to 1 : 50 000
(Pasajes, Bolivia). More precise coordinates were not
available for these test sites.
In order to compare the coordinates a transformation
equation has to be found. In our case this is done by a rel-
atively cost and time consuming Ground Control Point
(GCP)- method. Either the strict geometrical model - cen-
tral projection within the scan line and parallel projection in
flight direction - is applied using stereoscopic imagery, or
simple polynomials of first or second order are used which
are necessary if no stereo imagery is available.
In the first case the program BLASPO is used which per-
forms bundle adjustments for line scanners like SPOT or
MOMS-02 (Jacobsen, 1994). It utilizes neither actual and
precise orbit data nor automatically matched conjugate
points, but standard orbit values as well as additional
parameters for the affinity and angular affinity. Because
the mathematical model is based on the transformation
between two strictly orthogonal coordinate systems the
control points have eventually to be transformed from a
map projection (like UTM) into a local tangential system.
By theory one yields one set of exterior orientation param-
eters for every single scan line, but due to the high correla-
tion only every n-th line (n = 100 ... 10 000) has to be
considered.
The resulting transformation equation is then applied to
some Independent Control Points (ICPs) and compared to
reference coordinates. Or an entirely rectified image is
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
