International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
point targets is based on digital image processing techniques
and approximations from the manual measurements (Fig. 5).
Fig. 5: Sample image of the calibration target (left) and the
result of an automated detection of one point target (right)
The full set of image coordinates measured from 84 images
consists typically of about 14 000 positions within the set of
panchromatic cones and even more for the multispectral cones
of UltraCamD. The precision of the target detection algorithm
was investigated and a deviaton of about +/- 1.3 pm (i.e. 0.14
pixel) has been observed (Fig. 6).
2.0E DX (ym) | j
1.5 F f fie A ^u i . \ ; E
10 [ A A J hM |i Wu v ! 'N | \ :
asp ores NS 3
| s 50 100 150 200 250
2.0 E DY (ym) j
1.5F | M 4 À IM Jr 1
^E Il | A Y Y M M | 3
te i A lv | | FALE, iot ]
E 0 50 100 150 200 280
Fig. 6: Deviation of measured image coordinates of 200 target
points in the set of 84 images in x-direction (top) and y-
direction (bottom)
4. ESTIMATION OF CAMERA PARAMETERS BY
MEANS OF LEAST SQUARES ADJUSTMENT
The adjustment of all image coordinate measurements and the
estimation of camera parameters is performed with the software
package BINGO. The entire adjustment procedure is fourfold:
e Calculation of the initial solution of camera parameters
including parameters of CCD position, principal distance
and principal point coordinates and lens distortion.
e Transformation of image coordinates (the measurements)
to clear CCD position parameters. This step needs several
iterations in order to avoid any eccentricity of the radial
distortion parameters of the lens cone.
e Description of remaining displacements in the image plane
and description of the displacements by means of a look up
table.
e Estimation of transform parameters between cones in order
to guide the post-processing (stitching) of the large format
panchromatic image and the registration of the
multispectral channels to the panchromatic image
(pansharpening).
4.1 Results
The adjusted parameters and the look up tables are stored
together in a calibration data set. This dataset is automatically
used during the postprocessing by software. The final output
image (production image) of the UltraCamD is distortion free,
the remaining set of camera parameters which need to be
known by the user consist of the well known parameters of a
pinhole camera, namely the principal distance and the
coordinates of the principal point.
The standard deviation of these parameters is recalculated after
the calibration process and expected to be less then +/- 2 um.
5. CONCLUSIONS
The multi-cone/multi-sensor concept of UltraCamD demands a
specific calibration method. This calibration method is based on
a highly redundant set of image coordinate measurements and a
least squares solution of the well known bundle adjustment
technology. New parameters have been introduced in the
adjustment software package BINGO to model UltraCamD, its
cones and lenses as well as the set of 13 CCD sensors.
The output of the adjustment is a precise description of the
camera geometry, including CCD position, principal point
coordinates, principal distance and lens distortion parameters.
The accuracy of the calibration method was analyzed by
exploiting the flexible stochastic model offered by BINGO. To
quantify the result by a single number we have used the
sigma o value after the adjustment. This value has been‘
detected to be in the range of +/- 1.0 um to +/- 1.5 um for all
calibration session done so far.
6. REFERENCES
Kruck, E. (1984) BINGO: Ein Bündelprogramm zur
Simultanausgleichung für Ingenieuranwendungen =
Möglichkeiten und praktische Ergebnisse, Intenational Archive
for Photogrammetry and Remote Sensing, Rio de Janairo 1984
Leberl F., R. Perko, M. Gruber, M. Ponticelli (2002) Novel
Concepts for Aerial Digital Cameras. ISPRS Archives, Volume
34, Part 1, Proceedings of the ISPRS Commission I
Symposium, Denver, Colorado, November 2002.
Leberl, F. et al. (2003): The UltraCam Large Format Aerial
Digital Camera System, Proceedings of the American Society
For Photogrammetry & Remote Sensing, 5-9 May, 2003,
Anchorage, Alaska
KEY WO
ABSTRA:
A discrete
shift, etc.
this paper,
mutually [
of the figu
of the resu
In image ı
the other 1
coordinate
extent the
influence :
low-cost
coordinate
function o
the result i
the methoc
(see, for e
Medioni a
of the fini
Pratt, 197
estimate tl
rectangula
the axes |
superpose:
estimation
Lebedev, :
lengths of
We assum
integer va
pointed pl
fis.1). Let
(0 1s a bac
object is €
vertical ai
element. T
elements (
the object
(according
contour ne
eight-conn
four-conne
lying betw
called zo
