amatic er-
data sets
nd 3 were
in reverse
h the hor-
lue of the
ystematic
turn table
‘y of each
ta sets =
er refined
nterval of
selection
cessary to
vershoot-
sor. After
to be the
jore than
y directly
surement
Is a time
| the turn
| Was au-
ment (15
chieve an
rval with
, the on-
h the fol-
iin" tele-
650 mm
length is
ne object
lerances.
tter A/8
is of the
n of the
Juced by
ompared
artificial
the final
he over-
uipment
collima-
(“Knife
collima-
stematic
angular
led by a
ancoders
angle is
‘the HR
> overall
iccuracy
e design
Optical Benches |n view ofthe required verification accura-
cies (sub-arcsec), the entire test setup (collimator, turn table,
MOMS-2P instrument) was mounted on a single rigid optical
bench with air pressurized support and on a concrete solid
block being entirely separated from the rest of the building
The entire integration / test facility was operated in con-
trolled clean room environment (class « 100000), and con-
trolled temperature and humidity.
Dedicated Software (Sub-Pixel Interpolation) A dedi-
cated software has been written to interface the test equip-
ment with the MOMS-2P signal processing electronics, to
control the turn table position, to acquire the precise angu-
lar readings and to compute the gravity center of the point
spread functions imaged at the CCD's.
2.3 Results
From the four data sets camera parameters and correction ta-
bles were deduced. The camera parameters (see Table 1) de-
scribe the locations of the CCD-arrays relative to each other,
the correction Tables 2 and 3 show the differences between
the actual and the nominal position of the sensor elements.
The positions are given with respect to a reference pixel co-
ordinate system, defined in the (object side) image plane at a
distance of the mean focal length FL from the projection cen-
ter. Its origin is at the center of pixel number 4501 of the high
resolution CCD-array HR5A. The positive y-axis is pointing in
flight direction, the positive x-axis along line in the direction
of decreasing pixel numbers (see Figure 4). Table 1 contains
[i I | ST 7 | 6000 | |
x + tv TSHR 11) 9600 |
LAN ST 6 :6000|
V Y (flight direction)
Figure 4: Reference pixel coordinate system
the mean focal length FL, the offset pixel numbers Nx
and the stereo angles vy for each channel. Nx,;; indicates
at which pixel number the y-axis intersects the CCD-array.
[ Channel | FE [mm] | Nxo;; | 71°] |
MS1 220.163 | 3006.0 0.00
MS2 220.096 | 3008.0 0.00
MS3 220.116 | 3016.8 0.00
MS4 220.088 | 30162 0.00
HR5 660.250 | 4500.5 0.00
ST6 237.241 3009.0 | 21.45
ST7 237.250 | 3020.1 | -21.45
Table 1: Calibrated camera parameters
93
In the Tables 2 and 3 the differences
daz d ZCAL -TNOM
dy YCAL — YNOM
between the calibrated and the nominal pixel positions are
listed. { real } and { ENOM } are derived for each
YCAL YNOM
calibrated pixel N according to equations 4 and 5:
XI E 1 Vsetl TX set?
YI iz 2 Yset1 + Yset2
TII rd Tset3 + Tset4 (3)
Yır 2 Yset3 + Yset4
TCAL lk] en Ar Í
EE (4)
YCAL 2 yr o Vri
TNOM | VS NN (5)
you | 2| Zitan(y)
with px — 10 um (pixel size).
—
Do
_
UII
derived from two data sets each. Their differences give an
indication how far the results are reproducible. The last two
lines in the Tables 2 and 3 give an average ¢ of these differ-
ences from all calibrated pixel positions of each channel. For
channels MS4, HR5 and ST7 the average difference ¢ exceed
the required 0.1 pixel, i.e. the relative locations of these CCD-
arrays were not determined with the required accuracy and
must be verified by inflight calibration. All CCD-arrays con-
sist of 6000 sensor elements. The HR5 channel is optically
combined by the two CCD-arrays HR5A and HR5B, which
overlap within a 3000 pixel interval. Both CCD-arrays were
calibrated separately. In Table 3, as in practical application.
they are treated as one single CCD-array with 9000 sensor
elements, consisting of HR5A-pixels 1 to 4500, correspond-
ing to HR5-pixel 1 to 4500 and HR5B-pixels 1501 to 6000,
corresponding to HR5-pixel 4501 to 9000.
The pixel position { i } and { e } are independently
yr
3 INFLIGHT CALIBRATION
The principle of inflight calibration can be considered as
the reverse process of photogrammetric point determination.
Both, inflight calibration and point determination are based
on photogrammetric bundle adjustment and require image co-
ordinates of homologous points and navigation data. While
point determination aims to extract object information using
the exact knowledge of the camera geometry, inflight calibra-
tion tries to extract information about the camera geometry
using exact knowledge of the object. Thus, a substantial
amount of very accurate ground control information is re-
quired. Most of the MOMS-02 stereo evaluations suffered
from a lack of sufficient accurate ground control points (GCP)
which are well identifiable in the imagery. This high resolu-
tion evaluation can only profit from control information which
has an accuracy of 5 m or better.
The geometric MOMS-2P inflight calibration will be real-
ized in cooperation of 3 institutions: The German Aerospace
Research Establishment (DLR), the Chair for Photogram-
metry and Remote Sensing, Technical University Munich
(LPF/TUM) and the Institut Cartografic de Catalunya (1CC).
For MOMS-2P the LPF/TUM and the ICC also cooperate in
a pilot project to map the whole state of Catalonia. Since
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996
