International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B6. Istanbul 2004
4.2.2 Platform calibration
The platform calibration is essential for the resampling of the
new large format image composite based on the four PAN
channels. Due to the fact, that a mechanical part used in high-
dynamic environments like a photogrammetric flight never can
be realized as absolutely stable, the DMC camera housing was
designed to allow for some angular deformation of the
individual camera heads relative to each other. These
deformations are different for each airborne environment and
have to be estimated from the mission data itself. This on-the-
fly calibration approach is based on tie point measurements
from the overlapping regions of pan-chromatic imagery.
Besides that, the precise knowledge of relative positions of the
individual camera heads, the calibration parameters from single-
head calibrations as described above and first approximations
on the relative misorientation between the camera heads are
necessary input data required for platform calibration. The
calibration is solved within a bundle adjustment approach,
where three already mentioned rotation angles plus a focal
length refinement for three camera heads relatively to one
reference camera head are estimated As mentioned in Dórstel et
al (2003) about 30-50 tie points are sufficient to estimate the
unknown parameters. The typically obtained accuracy is
reported with 1/12 to 1/6 of a pixel.
4.3 Leica ADS40
In contrary to the frame based approach (single or multi-head)
described so far, multiple linear CCD lines are used in the Leica
ADS40 system. The ADS sensor development was driven by the
experiences with digital airborne line scanning systems at DLR,
namely the WAOSS/WAAC camera systems, originally
designed for the 1996 Mars mission and adopted for airborne
use after failure of the mission. First tests with ADS engineering
models started in 1997, the official product presentation was
done during the ISPRS 2000 conference in Amsterdam. The
imaging part of the sensor consists of typically 10 CCD lines
with different viewing angles and different multi-spectral
sensitivity. Each individual line provides 12000 pix with 6.5 x
6.5 um? pixel size.
During calibration the pixel positions of each individual line are
determined. The nomenclature for the different CCD lines is
like follows: pan-chromatic forward (PANF), nadir (PANN)
and backward (PANB) lines, multi-spectral forward (red REDF,
green GRNF, blue BLUF) and backward (near-infrared NIRB)
lines. The viewing angle relative to the nadir looking direction
is also specified by extending these identifiers with the
appropriate inclusion of numbers representing the individual
viewing angle. For example 28 corresponds to the 28.4deg
angle between nadir and forward looking direction of the PANF
channels - the resulting identifier is PANF28. The other viewing
angles are 14.2deg for the backward PAN lines, 16.1deg for the
RGB forward lines and 2.0deg for the NIR backward looking
CCD line, resulting in 14, 16, 02 code numbers. Since each
PAN channel consists of two individual lines, shifted by half a
pixel (so-called staggered arrays), this two lines are differed by
using character A for the first and B for the second line. For
reasons of completeness it should be mentioned, that the ADS is
available with a slightly different CCD-line configurations in:
the focal plane also: In this case the nadir looking PAN
staggered lines and the forward looking RGB lines are
exchanged, resulting in nadir viewing RGB channels and an
additional forward looking PAN channel. Such configuration
might be advantageous, when the main focus of applications is
laid on the generation of MS ortho-images.
208
4.3.1 Lab calibration
The lab calibration of the ADS sensor is based on a coded
vertical goniometer (CVG) available at SwissOptic (a Leica
Geosystems company). All details on the calibration facilities
are given in Pacey et al (1999). The CVG was developed from
a modified electronic vertical goniometer (EVG), where the
photomultiplier is replaced by a digital CCD frame camera and
the glass reference plate (with its high-precisely known marks)
is replaced with a special glass code plate. These coded targets
are located at the two diagonals and the two horizontal and
vertical bi-sections of the plate. The spatial distance between
neighbouring targets is 10mm. The measurement is done
automatically with high precision. From the measured
corresponding object angles the calibrated focal length and the
distortion function are obtained. The CVG is used for the
calibration of classical RC30 cameras as well as for the ADS
sensors, although for ADS the calibration procedure has to be
modified like follows.
As described in Pacey et al (1999) lens cone and CCD focal
plate are calibrated separately first. Afterwards both
components are assembled and calibrated using the CVG. In
this case the glass code plate cannot be used any longer since
the CCDs are fixed in the focal plane now. Therefore, a coded
target is projected in reverse direction on to the CCD-line of the
tested lens. In order to allow measurements in of-nadir
directions an additional mirror scanner is mounted on top of the
goniometer arm. With this modification each individual pixel
location on the focal plate can be addressed. As written in
Schuster & Braunecker (2000) it is sufficient to measure pixels
every 2-5deg within the field of view. The values for
intermediate pixels are interpolated numerically.
4.3.0 Self-calibration by bundle adjustment
Although a complete measurement and process flow was
established for lab calibration a new approach for ADS
calibration was introduced recently. This in situ approach is
exclusively based on self-calibration, which is — as already
mentioned before — a system driven approach including the
calibration of all image-relevant system components. In this
context especially the inertial measurement unit (IMU) has to be
mentioned, which is essential for operational processing of
airborne line scanner data. The mandatory relative orientation
between IMU body frame and ADS photo coordinate system
can only be determined via self-calibration, which is one
advantage compared to the lab calibration approach. The
applied procedure is given in detail in Tempelmann et al (2003)
and should be recalled in the following.
The calibration is based on the orientation fixes approach
proposed by Hofmann, which is implemented in the bundle
adjustment software. Again the Brown parameter sets are used
as calibration terms. Beside that, additional three unknowns are
used to model the before mentioned misalignment angles.
Although ADS40 comprises line instead of classical frame
geometry, many of the Brown parameters are directly
transferable. Some of the parameters (modelling platen flatness)
are not useful for line scanners and have to be eliminated.
Nonetheless, some uncompensated effects remain. These
remaining effects, which are non compensated via the Brown
parameter set, have to be modelled by additional polynomials.
In Tempelmann et al (2003) a 6" degree of order polynomial
performs sufficiently well and is recommended for X and Y
components of each sensor line. This extended model will be
available in the updated bundle software, hence additional
polynomial coefficients are directly estimated in the bundle.
In order to realize a sufficiently well overall system calibration,
special requirements for the calibration flight pattern are