‚us to
Also,
t only
d and
a and
lively
le for
ilding
turns,
ns on
edge
faces
e and
esults
urther
hieve
4-25.
ilding
ation
SPRS
l and
(eds),
roject
3, pp.
erlag,
on of
n Ill
mbus,
SPRS
Manfred Weisensee
COMBINED ADJUSTMENT OF LASER SCANNING DATA
AND DIGITAL PHOTOGRAMMETRIC IMAGES
Manfred WEISENSEE
Fachhochschule Oldenburg/Ostfriesland/Wilhelmshaven, Germany
Institut für Angewandte Photogrammetrie und Geoinformatik
weisensee @ fh-oldenburg.de
Working Group IC-15
KEY WORDS: Data fusion, Digital Terrain Model, Fast Vision, Laser Scanning, Modeling, Orthophoto,
Photogrammetry, Surface reconstruction.
ABSTRACT
In this paper, a method is described, which introduces data from laser scanning devices into a process of automated
photogrammetric surface reconstruction. The results of this process are the ortho image, the object surface that
conforms with the ortho image and the areas, where laser data have been excluded from the reconstruction process.
First, the concept of Facets Stereo Vision - a general method of automated photogrammetry - is explained and an
efficient process of surface reconstruction is given. Then, laser scanning data is introduced as additional observations
into the reconstruction process.
1 INTRODUCTION
Although the determination of object surface geometry has reached a high degree of automation using state-of-the-art
methods of digital photogrammetry, in a very short time, laser scanning systems have already surpassed
photogrammetry in some topographic as well as close-range applications. Considering the different fields of application
of both techniques in geometric and semantic data collection, three important tasks can be outlined:
1. the production of true ortho images, which require a surface model including vegetation and man made objects like
buildings and bridges (upper surface),
2. the collection of digital terrain models, usually in a regular spacing with additional information, e.g. breaklines, for
various applications like contour interpolation, which require true ground information (lower surface) instead and
3. the discrimination of objects from their surrounding for mapping and monitoring purposes, which can be seen as
the difference model between the two surfaces described above.
Both techniques are able to collect a large amount of geometric data in a very short time, but they also have in common,
that a remarkable amount of user interaction is still necessary to verify, edit and refine that data, while the extraction of
semantic information in either way to a large extent still is carried out by operators interactively. The reasons for user
interaction even during or after geometric data collection due to failures or deficiencies of the methods are numerous:
partially or totally occluded regions in forests or built up areas, insufficient reflectance of the surface leading to faint
laser or image signals, multiple reflections due to surface roughness, moving or even flying objects and many more.
Numerous proposals have been made to overcome some of the deficiencies especially of laser scanning, e.g. the case of
collecting the lower surface in forests. Some of the catchwords are: robust estimation, linear prediction or
morphological filtering, cf. Pfeiffer et al. in: /Steinborn and Fritsch 1999/, /Lindenberger 1993/. Almost any of the
proposed methods would benefit from the knowledge of the upper surface and also some information about areas,
where a lower surface has to be expected. In this paper, a method is described, which introduces data from laser
scanning devices into a process of automated photogrammetric surface reconstruction. The results of this process are the
ortho image, the upper surface and the areas, where laser data have been excluded from the reconstruction process.
2 PHOTOGRAMMETRIC SURFACE RECONSTRUCTION WITH FACETS STEREO VISION
The concept of Facets Stereo Vision (FastVision) has been published first in /Wrobel 1987/. Instead of matching small
patches of one image to corresponding patches of another image by applying geometric and radiometric transformations
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 965