the surface discontinuities. The breaklines and the laser data are
then merged to create a new data set.
The most significant surface discontinuities occur due to
buildings, thus the rooflines of buildings are collected and used
for the data fusion. As only the rooflines are measured, the
surface at the ground level will not be adequately defined.
Therefore, to better define the visible surface, it is necessary to
incorporate measurements at ground level surrounding the
buildings. This is accomplished by vertically projecting the
horizontal position of the roofline to ground level. The elevation
of the ground at that point is estimated using the surrounding
laser data points. The breakline from these new points is slightly
offset from the position of the roofline, as the triangulation
procedure does not accommodate points with exactly the same
horizontal position.
4.1 Surface Generation and Registration
Initial surfaces were generated to enable the registration of the
two data sets. A digital elevation model (DEM) was generated
using the softcopy photogrammetry program OrthoMax, under
the Erdas Imagine environment. The laser surface used for the
registration process is shown in Figure 2.
Figure 2. Laser surface used for registration.
The transformation parameters were determined using the
developed registration algorithm. As a check on these results,
roof breaklines were measured analytically to produce planes, and
the laser points which occurred in the vicinity of these planes
were used in the registration process, therefore providing a result
which was not dependent on automatic DEM generation
techniques. The results were of similar magnitude in each case,
with the vertical shift between the surfaces being approximately
one meter. Further details are provided in Postolov et al. (1999).
4.2 Results of Data Fusion
For the current experimentation, the breaklines have been
measured manually using an photogrammetric workstation. Roof
breaklines were collected over a residential area in the stereo
model. The laser points are filtered to delete points which fall on
breaklines or within a certain distance of the breaklines. A small
area is shown in Figure 3 to more clearly present the information.
The same area is presented in the following figures. The surface
generated from the laser data is shown in Figure 4.
International Archives of Photogrammetry and Hemote Sensing, Vol. 32, Part 3W14, La Jolla, CA, 9-11 Nov. 1999
Figure 3. Breaklines and laser points.
A new surface is generated using the merged data. Constrained
triangulation is used to enforce the use of the breaklines. The
surface using the merged data is shown in Figure 5. Comparing
these surfaces, the merged surface provides a better
representation of the buildings than the surface generated using
only the laser data.
The breaklines used up to this point only define the rooflines. To
more accurately define the buildings, the roofline is projected on
to the ground surface and used as a breakline, thus defining the
walls of the buildings in addition to the roofs. The result of this
process is shown in Figure 6.
Comparing the surfaces, the surface utilizing both the breaklines
and the laser data better defines the buildings than either of the
surfaces generated using a single data acquisition method, and
that the projection of the roofline to the ground surface is
necessary to accurately define the buildings.
5 CONCLUSIONS
This paper describes the development of a general scheme for the
integration of laser and photogrammetric data. This scheme
includes the development of a surface registration algorithm and a
data fusion algorithm. Initial testing of the integration algorithm
has been carried out using data over Ocean City, MD, and has
shown that the surface is more accurately represented than when
using either data set separately.
The approach presented in this paper is applicable to the
determination of accurate visible surface models for the
generation of true orthophotos in urban areas. The use of the
rooflines to determine surface discontinuities will allow the
production of an orthophoto that does not have the distortions
associated with the inaccuracies in the surface models which are
inherent in automatically generated DEMs. Continued research is
being undertaken into the automation of the approach, to extract
the edges of the rooflines and to determine the ground height in
these regions.
ACKNOWLEDGEMENT
This research project is partially supported by the United States-
Israel Binational Science Foundation, grant no. 97-00433.
0 4670 4580 4690 4700 4710 410 ——
Internation
REFERENCES
Ackermann, F., 199€
and future expectatio
Remote Sensing, 54(1
Axelsson, P., 1998.
and topographic dai
Photogrammetry, Ist:
Axelsson, P., 1999. I
and applications. ISP
Sensing, 54(1), pp. 1.
Baltsavias, E., 1999.
laser scanning. /SPR
Sensing, 54(1), pp. 8.