point
> and
2359
COOr-
enta-
ation
lades
eters
Be-
focal
coef-
1 the
ACCU-
n by
notes
f the
(po-
in à
| per-
rigin
rallel
idian
The
ntrol
pre-
n de-
lages
(atti-
; the
er to
just-
) the
ional
val-
GCP
0m)
(o=
01°
nma-
Was
achieved. The rms value u$; of the theoretical stan-
dard deviations of the adjusted object point coordinates
amounts to 180 m, which gives a measure of the interior
accuracy of the block. The exterior accuracy of the block,
however, is in the order of 4 km (= mean absolute accuracy
of the adjusted position parameters).
| Value |
60 [pm] 7.8
wx [ml | 289
u$ [m] 99
m [m] 64
MxYz [m] | 180
Table 1: Results of the bundle block adjustment
In addition, the coordinates of 33 object points have
been compared with those, which were determined in-
dependently by M. Davies, RAND Corp., Santa Monica,
USA (Davies et al. 1995). Both control networks corre-
spond well, as the rms value of the coordinate differences
paxayaz=200m indicate.
In a separate processing step, which has been not yet real-
ized, the existing bundle block adjustment will be supple-
mented by a rigorous dynamical modeling of the spacecraft
motion to account for orbital constraints (Montenbruck et
al. 1994, Ohlhof 1996a). This advanced concept ensures
the proper utilization of Galileo trajectory information in
the bundle block adjustment and, vice-versa, allows the
use of image information to improve the orbit determina-
tion and supports the estimation of dynamical parameters,
e.g. the rotational parameters of Ida or the ephemeris of
Ida’s satellite Dactyl.
The incorporation of orbital constraints into the bundle
adjustment has been first realized and successfully applied
to practical MOMS-02/D2 and simulated HRSC/WAOSS
Mars96 multi-line imagery (Ohlhof 1996b).
Based on a theoretical analysis of the Galileo spacecraft
trajectory, Gill et al. (1995) have been found out, that a
simple linear model of the spacecraft position w.r.t. time
describes the orbital motion quite well. Within a 30 min
interval around encounter time the deviations from the
linear model due to Ida’s gravitational field stay below
25 cm for the position and 0.1 mm/s for the spacecraft
velocity, that was below the precision of the ground-based
S-Band Doppler observations (2 mm/s).
2.3 DTM Generation
Third, a Digital Terrain Model (DTM) is generated. The
DTM covers one hemisphere of Ida with a resolution of
1° x 1°, which can be interpreted as a shape model of the
asteroid. It uses a sphere as reference surface.
DTM generation involves the determination of a large
number of conjugate points in the images, the computation
of ground coordinates for these points and the approxima-
tion of the object surface. Digital image matching is a
suitable technique to find the required number of conju-
gate points automatically. Using the least squares region-
growing matching algorithm (Otto, Chau 1989) about
32,000 image points on Ida’s surface were found in 10 im-
623
ages having 30-110 m ground pixel size. The lower reso-
lution (< 110 m ground pixel size) images of Ida are not
suited for image matching, so that the conjugate points
cover only one hemisphere of the asteroid. The tie points
determined previously by the human operator are utilized
as starting (seed) points for the matching procedure.
Special methods were developed for automatic point trans-
fer in multiple images and for the consideration of scale dif-
ferences between the images up to factor 3.5. In addition,
4,200 matched image points in deep space were deleted au-
tomatically as blunders using a given treshold value. After
that, the computation of ground coordinates was carried
out via forward intersection using the adjusted exterior
orientation parameters.
The ground coordinates were processed using an approx-
imation method for scattered data on a sphere (Brand et
al. 1995; Brand, Frohlich 1996). The height information
of irregular distributed data points is transferred to a reg-
ular grid by calculating weighted means in a spherical cap
around a grid point. Repeating this calculation for dif-
ferent radii of the spherical caps, the irregular data can
be handled adequately. In a hierarchical algorithm these
calculations are only done in regions in which the error
is above a threshold. Note, that the radius of the spher-
ical cap, which is comparable with the mesh size of the
usual planar approaches, is fitted to the resolution of the
data. The 1? x 1? DTM, which is visualized in Figure 3,
was calculated from the regular grid using the smoothing
technique described above.
In Figure 3 an illumination from the upper left direction is
assumed and the DTM is shaded using a Gourand shading
algorithm. The part of Ida image s0202561900 presented
in Figure 4 can be found on the upper left side of the
DTM in Figure 3. Due to the scale of the DTM and the
distribution of the DTM primary data, local features such
as craters cannot be represented.
Figure 3: DTM of Ida using a reference sphere
With the help of digital terrain models, color orthoimages
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
