Figure 6 The new ITC. 
Analytical and digital photogrammetry can deliver 3D 
topographic data. Although measuring by photogrammetry 
is traditionally done in 3D, the transition from 2D to 3D 
models considerably increases the effort needed for data 
collection and data structuring. It also drastically expands 
data volumes. Manual extraction of houses does not pose 
fundamental problems, only practical ones. Until further 
progress is made in semi-automatic extraction, strict 
digitizing rules should be followed to minimize measuring 
866 
time and facilitate automatic 3D model construction, 
especially in cases of complex buildings and when a high 
level of detail is required. We have not yet accomplished 
fully automatic model construction, since building of 
topology is significantly more demanding in 3D than it is in 
2D. Stereosuperimposition of 3D wireframes is useless if 
there is no hidden line removal, since it becomes 
indecipherable. It is also useless for interactive correction if 
the digitizing software does not support 3D topology. 
Conventional stereoplotting software can only keep the roof 
polygon 'alive' when we move a corner point, but cannot 
modify the walls accordingly. Attempting to change a corner 
point of a footprint would be an even more insane 
undertaking since the coding system only recognizes roofs, 
thus even off-line rebuilding of topology is not possible. 
Current DPs offer a higher level of computer support than 
APs, but this does not concern feature extraction for 
topographic mapping. Our comparative tests using the C120 
and the T10 showed no significant differences in time 
efficiency, which conforms with the results reported by 
(Colomer et Colomina, 1994). This implies that the 
additional costs of working all-digitally cannot be recovered 
yet, unless a hybrid analytical-digital production line is set up 
for 3D data collection or a significant productivity boost is 
attained through automating feature extraction. 
5. FURTHER DEVELOPMENT 
Digital photogrammetry offers the prospect of automated 
information and texture extraction from photographs; in 
both it is distinct from analytical photogrammetry. Semi- 
automatic, local image matching on digital plotters seems 
feasible on short term, thus speeding up reconstruction of 
certain objects such as roads and buildings. Once the 
geometry of the topographic objects is known, the 
candidate areas for texture mapping can automatically be 
extracted form the digital images. 
Apart from (semi-)Jautomating object recognition and feature 
extraction, system enhancements are also needed, 
especially for efficient quality control and updating 
operations. Such desires require very powerful graphic 
subsystems with large image memory for effective viewing 
of object reconstructions. Because of the possibility of 
occlusions in built up areas and the countless number of 
possible object shapes and compositions, only a direct 
visual comparison of object reconstruction with the original 
images can give a final validation. Efficient verification would 
be possible if an object could be viewed--almost 
instantaneously--in more than one stereomodel. Adequate 
3D viewing must be combined with interactive digitizing 
software that supports 3D topology. With fully automatic 
extraction of topographic information from aerial 
photographs still miles away, extended image access and 
computer graphics (integrating digital photogrammetry and 
CAD) will increase productivity of 3D vector data acquisition 
and thus make a broad introduction of 3D-GIS more 
feasible. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996 
NI << N 
<< DD = P= == po