SSI image data Navigation data
{À
[NU ®
Image
Orientation
Measure tie points
Convert pixel coordinates
to image coordinates
|
|
Combined bundle block adjustment
SSI Ground Improved
image data control exterior
network otientation
DTM
Generation
Determine
conjugate points
I
Convert pixel coordinates
to image coordinates
Forward intersection
Surface modelling
Digital terrain
model
Figure 2: Data flow for photogrammetric processing
measurement of tie points as well as the combined block
adjustment and provides improved exterior orientation pa-
rameters and a 3-D ground control network on Ida. For the
acquisition of DTM primary data, digital image matching
is applied leading to a large number of conjugate points in
the images. With the help of improved exterior orienta-
tion parameters the image coordinates of conjugate points
can be transformed into ground coordinates using forward
intersections. Finally a DTM is generated using a sphere
as reference surface.
2.2 Image Orientation
The first photogrammetric processing step involves the de-
termination of a couple of tie points in the images. Point
identification and point measurement were carried out by a
human operator. All points are related to topographic fea-
tures (mostly craters) on Ida’s surface and are distributed
uniformly on the ground surface in order to build up a
dense global network.
Altogether 96 points were measured interactively in 36
images with a precision of 10 um (0.7 pixel), where the
ground pixel size ranges from 25 m to 600 m. The point
identification was impaired by the varying image scale and
by the fact that the asteroid rotated on its spin axis 235?
during the time period covered by the images.
The following step is to convert the measured pixel coor-
dinates into image coordinates taking the interior orienta-
tion into account. To this end, a geometric star calibration
was carried out by measuring star locations of the Pleiades
star cluster in the images and solving for free parameters
in a least-squares adjustment (Davies et al. 1994). Be-
sides the 3 image orientation angles, the calibrated focal
length c — 1500.467 mm and the radial distortion coef-
ficient k = —0.00002498 mm”? were determined in the
least-squares adjustment in the order of 0.1 pixel accu-
racy. This simple model is sufficient for images taken by
narrow-angle cameras like SSI.
The pixel coordinates (row,col) can be transformed by
rz = ps(col — 400.5) y = ps(row — 400.5) (1)
ds = z(1+ kr’) uy = —y(1+ kr’) (2)
with
7? = 1° + y? (3)
where ps is the pixel size (15.24 um) and (uz, uy) denotes
the image coordinates.
In the next processing step, 3D ground coordinates of the
tie points are determined and the exterior orientation (po-
sition and attitude) of all images is reconstructed in a
bundle block adjustment. The block adjustment was per-
formed in the Ida-fixed reference system, where the origin
coincides with Ida's center of mass, the Z-axis is parallel
to the spin axis, and the X-axis lies in the prime meridian
plane defined by the crater Afon (Davies et al. 1995). The
crater Afon was introduced as error-free ground control
point (GCP) for the definition of the global datum.
Additional observation equations are formulated for pre-
processed position and attitude data, which have been de-
rived from S-Band Doppler tracking data and star images
respectively. The relative accuracy of the position (atti-
tude) data is assumed to be 100 m (0.1°), whereas the
absolute accuracy amounts to 5 km (0.2°). In order to
incorporate these navigation data into the bundle adjust-
ment, they are transformed from inertial space into the
Ida-fixed non-inertial coordinate system. The rotational
parameters of Ida are treated as constants using the val-
ues of Davies et al. (1995).
The following data were introduced as observations:
e Image coordinates of 95 conjugate points and 1 GCP
(c —10 um)
e Object coordinates of 1 GCP (ox =0y=0z=10 m)
e Position parameters (z^, y^, z^) for 36 images (0 =
100 m relative, c —5 km absolute)
e Attitude parameters ($, v, K) for 36 images (g=0.1°
relative, g=0.2° absolute)
The results of the bundle block adjustment are summä-
rized in Table 1. After 6 iterations a Go of 7.8 um was
622
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
Cy ^" /G c3 rj, OD
CD cH co cho co do LN oM M MH wa N — 0, rm rM
— mM
eR t