ISPRS Commission III, Vol.34, Part 3A „Photogrammetric Computer Vision", Graz, 2002
lem, the surface can be considered to be known. Hence,
the problem of geometric distortions of features can be well
controlled and matching becomes feasible, assuming that
reasonable approximations of the exterior orientation pa-
rameters are available. In fact, matching in object space,
using iteratively warped images, becomes the method of
choice. This matching procedure also offers the opportu-
nity to match multiple images. Now we have the chance for
a detailed reconstruction of complex surfaces from multiple
images.
Orientation with surface patches Originally, the idea of
using surfaces in the form of DEMs for orienting models was
suggested by Ebner and Strunz (1988). The approach is
based on minimizing the z differences between the model
points and points in object space found by interpolating the
DEM. The differences are minimized by determining the ab-
solute orientation parameters of the model. Schenk (1999a)
modified the approach by minimizing the distances between
corresponding surface elements. We propose the latter
method for fusing aerial images with LIDAR.
The advantage of using patches as sensor invariant features
is the relatively simple process to extract them from laser
points. The fitting error of the laser points to a mathematical
surface serves as quality control measure. In contrast to
the previous methods, no planimetric features need to be
extracted. Patches are more robust than features derived
from them, for example 3D edges.
Unfortunately, the situation is quite different for determin-
ing surface patches in aerial images. Although theoreti-
cally possible by texture segmentation and gradiant anal-
ysis, it is is very unlikely that surface information can be
extracted from single images. Hence, image matching (fu-
sion) is required. Quite often, surface patches have uniform
reflectance properties. Thus, the grey level distribution of
conjugate image patches is likely to be uniform too, preclud-
ing both, area-based and feature-based matching methods,
respectively. The most promising approach is to infer sur-
face patches from surface boundaries (matched edges).
As shown by Jaw (1999), the concept of using control sur-
faces for orienting stereo models can be extended to block
adjustment. In analogy to tie points, the author introduces tie
surfaces. To connect adjacent models, the only condition is
to measure points on the same surface. However, the points
do not need to be identical—clearly, a major advantage for
automatic aerial triangulation.
Alternative solution with range images A popular way
to deal with laser points is to convert them to range images.
This is not only advantageous for visualizing 3D laser point
clouds but a plethora of image processing algorithms can
operate on range images. For example, an edge opera-
tor will find edges in a range image, suggesting that the 3D
edges used as sensor invariant features be determined from
range images. At first sight, this is very appealing since it ap-
pears much simpler than the method described in Section 2.
Let us take a closer look before making a final judgment,
however.
Generating range images entails the interpolation of the ir-
regularly spaced laser points to a grid and the conversion of
elevations to grey values. While the conversion is straight-
forward, the interpolation deserves closer attention. Our
goal is to detect edges. Edges in range images correspond
A - 314
to rapid changes of elevations in the direction across the
edge. This is precisely where we must expect large interpo-
lation errors. It follows that the localization of edges in range
images may not be accurate enough for precise fusion.
Figure 2: Edge detection performed on a range image. The edges
are affected by interpolation errors and usually not suit-
able for sharp boundary delineation.
Fig. 3.3 depicts a sub-image with a fairly large building (see
also Fig. 3b). The DEM grid size of 1.3 meters (average
distance between the irregularly distributed laser points)
leads to a relatively blocky appearance of the building and to
jagged, fragmented edges. Moreover, edges are predomi-
nantly horizontal. It is quite difficult to detect non-horizontal
edges (boundaries) in object space from range images. Fi-
nally, when comparing the edges obtained from the range
image with those determined by intersecting adjacent pla-
nar surface patches it becomes clear that their use for fusing
aerial images with LIDAR becomes problematic.
4. FUSION OF AERIAL IMAGERY WITH LIDAR DATA
After having established a common reference frame for LI-
DAR and aerial imagery we are now in a position to fuse
features extracted from the two sensors to a surface de-
scription that is richer in information as would be possible
with either sensor alone. We have strongly argued for an
explicit description to aid subsequent processes such as ob-
ject recognition, surface analysis, bare-earth computations,
and even the generation of orthophotos. Since these appli-
cations may require different surface descriptions, varying
in the surface properties (quality and quantity), an impor-
tant question arises: is there a general description, suitable
for applications that may not even be known by the time of
surface reconstruction?
Surfaces, that is their explicit descriptions, play an impor-
tant role in spatial reasoning—a process that occurs to a
varying degree in all applications. We consider the surface
properties listed in Table 2 essential elements that are, by
and large, application dependent. In a demand-driven im-
plementation, additional properties or more detailed infor-
mation can be obtained from the sensory input data upon