You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Technical Commission VII

International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
U. Beisl*
* Leica Geosystems AG, 9435 Heerbrugg, Switzerland — Ulrich.Beisl@leica-geosystems.com
Commission VII, WG VII/1
KEY WORDS: Atmosphere, Modelling, Radiometric, Calibration, Processing, Multispectral, Digital, Camera
The image quality of photogrammetric images is influenced by various effects from outside the camera. One effect is the scattered
light from the atmosphere that lowers contrast in the images and creates a colour shift towards the blue. Another is the changing
illumination during the day which results in changing image brightness within an image block. In addition, there is the so-called
bidirectional reflectance of the ground (BRDF effects) that is giving rise to a view and sun angle dependent brightness gradient in the
image itself. To correct for the first two effects an atmospheric correction with reflectance calibration is chosen. The effects have
been corrected successfully for ADS linescan sensor data by using a parametrization of the atmospheric quantities. Following
Kaufman et al. the actual atmospheric condition is estimated by the brightness of a dark pixel taken from the image. The BRDF
effects are corrected using a semi-empirical modelling of the brightness gradient. Both methods are now extended to frame cameras.
Linescan sensors have a viewing geometry that is only dependent from the cross track view zenith angle. The difference for frame
cameras now is to include the extra dimension of the view azimuth into the modelling. Since both the atmospheric correction and the
BRDF correction require a model inversion with the help of image data, a different image sampling strategy is necessary which
includes the azimuth angle dependence. For the atmospheric correction a sixth variable is added to the existing five variables
visibility, view zenith angle, sun zenith angle, ground altitude, and flight altitude - thus multiplying the number of modelling input
combinations for the offline-inversion. The parametrization has to reflect the view azimuth angle dependence. The BRDF model
already contains the view azimuth dependence and is combined with a new sampling strategy.
Originally, photogrammetric camera systems were used for
metric purposes in the geometric domain, i.e. for measuring
distances, areas and angles. This was possible with the analog
film cameras which provided sufficiently good contrast and
sharpness. Attempts were made to use the wet film technology
for radiometric measurements using densitometers, but the
radiometric resolution was poor and the results were only stable
within one film roll due to the influences of temperature and
developer during film development.
With large format digital sensors becoming affordable for
photogrammetric users ten years ago, new application areas
have developed quickly. Remote sensing applications which
could be handled only with calibrated satellite images can now
be solved with airborne images, too. In addition, a new mass
market for cheap high resolution images for use in internet
based mapping systems has emerged. In addition to a minimal
geometric accuracy the new applications require a balanced
radiometry and removal of atmospheric artefacts.
When digital cameras appeared on the market the analog film
data workflow had already turned digital by using film scanners.
Therefore the geometric calibration algorithms could be easily
transferred to the digital image data workflow. The radiometric
processing of digital camera images had long been dominated
by a mere relative calibration of the lens falloff.
However, the large field of view (FOV) and the varying flying
height of airborne cameras introduce strongly varying effects of
atmospheric stray light, giving rise to a blue hue, increasing
towards the borders of the images. Furthermore the effects of
bidirectional ground reflectance (BRDF) cause varying
brightness within the image, the most prominent ones being
sunglint in the water and a hot-spot in the image at high sun
elevation. To address these radiometric aspects an EUROSDR
project was initiated (Honkavaara, 2011).
In order to correct those environmental artefacts in airborne
images, methods from satellite and hyperspectral airborne image
workflows were adapted to the needs of high-resolution
photogrammetric images. Those methods use physical models
which require an absolute calibration of the airborne sensors.
The Leica ADS40 camera was the first commercial
photogrammetric camera that provided an absolute radiometric
calibration (Beisl, 2006). This was the prerequisite for applying
an automated atmospheric correction in the photogrammetric
workflow, which was implemented together with a BRDF
correction (Beisl et al, 2008). The atmospheric correction
option for ADS image data has become the standard setting in
the image workflow for XPro users (Downey et al., 2010). A
validation of the reflectance calibration has been presented by
(Markelin et al., 2010) and (Beisl et al., 2010).
The Swiss Federal Office of Topography (swisstopo) is now
using absolutely calibrated ADS images to produce two quality
image products in an operational way (swissimage standard
product and remote sensing basis product) (Schläpfer et al,
This paper gives an outline, how to extend the ADS radiometric
correction algorithms for use in frame sensors like DMC (Ryan
et al. 2009) or RCD30 (Wagner, 2011), (Tempelmann, 2012).