endered,
jetail at
graphic
ists, one
llipsoid
a plane
ast with
s being
ain may
Earth is
objects
tries to
creasing
interest
y deve-
omputer
require-
eded to
es. The
ed to be
ved and
1easured.
) remote
tain the
nination
: surface
photo-
ons of a
is being
h.
omputer
ton but
ample is
be com-
ween a
s useful
n image
t al. (in
ages can
a digital
n actual
t to the
1ces be-
ed as a
nth sub-
| 60,000
fice. The
on have
ation of
. method
g serves
ty is a
1 a room
g ds de-
y is not.
has been
S a wire-
Figure 10.1: Photograph of a room.
Figure 10.2: Wire frame rendering of the geometric model of
the surface with ~ 60000 elements (from Karner, in print).
Figure 10.3: Rendering of the room using surface properties,
geometry and illumination via a radiosity model
(from Karner, in print).
11 IMMERSION IN 3-D SCENES
Surfaces are being perceived by a human computer
operator when wearing stereo glasses. The object that
is being viewed can be interacted with if a 3-
dimensional cursor is placed in the field of view. The
computer now needs to be able to track the position of
the cursor as well as the position of the stereo glasses.
As the cursor moves the computer is aware of those
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
motions and can establish a relationship between the
virtual 3-dimensional object and the real position of
the cursor. In this manner a user can edit a 3-
dimensional object interactively. The user can also
move freely with the head if the computer can track the
position of the head and can change the 3-dimensional
object’s visualization as a function of the viewer
position. In this manner a fully 3-dimensional object
could be inspected from the left or the right, from
above or below depending on the position of the
viewer's eyes.
This is denoted as ,iimmersion* in 3-d scenes and we
are describing an immersive user interface based on a
3-dimensional object surface and the 3-dimensional
cursor, accompanied by a tracked viewer. This concept
can be a significant element in a chain that goes from
sensor sources via point clouds to surfaces and
realistic visualizations. The immersive user interface
¢an support the quality control and editing in a fully
digital environment at very modest to no additional
equipment costs.
12 OUTLOOK AND CONCLUSION
In recent years we have seen a dramatic broadening of
the ideas of surface reconstruction. We have described
a generalization of these ideas from traditional photo-
grammetric stereo mapping of topographic surfaces to
the use of a whole range of techniques, all aiming at
geometric modeling of the surface of terrain or fully
3-dimensional objects. We have also shown a broaden-
ing from the traditional measurement of surface points
to the more modern integrated view of „digital visual
information“, to include the information source in the
form of images or raw surface measurements, to
combine geometry and radiometry and finally to
include the application in the visualization and
interaction with 3-dimensional digital surfaces.
The traditional problems of image matching have
made slow progress. The human operator is still the
best stereo measuring entity. The machine is capable
to perform measurements the human cannot take such
as ranging, operating with structured light, performing
interferometry or doing altimetry. The human can use
single image or stereo object reconstruction, not
however any of the other techniques. In a trade-off
between manual human-based surface reconstruction
and automated machine vision the evolution is to-
wards largely automated systems in which the human
is merely the quality control and editing agent. This
requires that efficient interaction with 3-dimensional
data be available in a form that is near a specific
application scenario. The geographic information sy-
stem GIS is in the process of generalizing from 2 to 3
dimensions, e.g. by converting the 2-dimensional
multi-layer urban GIS into a CyberCity model. The
requirements of planetary mapping, of automated re-
mote sensing, or of industrial inspection all point to-
wards an increased use of automated processes to
model the geometry of a surface and to provide tools
to also determine non-geometric properties.
We believe that these developments need an inte-
grating view that is capable of fusing various data
sources and analysis methods. One obtains a toolbox
of procedures so that human interaction is small and
limited to improvement of the quality of surfaces.