Figure 7. Perspective SW of the 3D model of the Oeiras mother
church, rendered with rectified images
5.6 3D Urban Model
The steps to create a 3D urban model are shown in Figure 2.
Data acquisition refers to the acquisition of the basic input, such
as points and lines describing the roofs of buildings and a
Digital Terrain Model (DTM). These are extracted/produced by
means of photogrammetric restituion of aerial photographs.
Modelling was done by surfaces with MicrostationJ. Figure 8
shows the basic information projected onto the DTM. The
surfaces representative of each wall are created afterwards.
Rendering concerns mapping of textures using a rendering
algorithm. To make the model realistic, the textures are made
with rectified images of the facades of the objects. The
rendering algorithm used allows to add to the model some
caracteristics such as ilumination, transparency and colour.
Figure 9 shows a part of the 3D model of the Ociras mother
church after rendering.
Basic Information
+
Projection onto the DTM
+
Surface Creation
Front (corner 1, 2, ...)
Right (corner 1, 2, ...)
Left (corner 1, 2, ...)
Figure 9. Part of the 3D model of the Oeiras mother church and
its surroundings
5.7 Visualisation
Visualisation is a complement to the architectural archive, as a
means of divulging it. In addition to the various images that
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004
may be extracted to visualise results, animations were created
with Microstation. These are walk-throughs around the church
and its surroundings, which were recorded on a CD, integrated
with other products such as the recorded facades, the 3D urban
model and information about the Oeiras church history.
5.8 Comparative Analysis- Stereoscopic Photogrammetry
versus Monoscopic Photogrammetry
The comparative analysis between the two photogrammetric
techniques was carried out by computing statistical values for
the residuals between the coordinates of the points measured by
the referred techniques and those of the control points. À total
of 100 control points were chosen randomly and their
coordinates measured with precise topographic techniques. The
points were measured on images acquired with the different
cameras (Table 2). The computed statistical values refer to the
Root Mean Square Error (RMSE), the standard deviation and
the maximum and minimum residuals (in X, Y and Z). Because
of apparent incompatibilities between the digital metric Rollei
camera and the photogrammetric workstation, which require
further investigation, Table 3 lists only the values computed for
the metric camera Leica RS and the stercoscopic technique.
Leica R5
X Y Z
(em) | (em) | (em)
RMSE 2,3 2,43 4,0
Standard Deviation | 2,1 2,4 4.0
Maximum Residual | 5,4 6,1 6,5
Minimum Residual -5,6 -5,2 -5,7
Table 3. Statistical results for the residuals obtained with a
metric camera and the stereoscopic photogrammetry
(ImageStation SSK Pro)
Tables 4a and 4b list the results for the monoscopic
photogrammetry using all the cameras, and Table 5 shows
which comparisons were carried out between the results
obtained with the different cameras and photogrammetric
techniques. These comparisons involved two statistical tests,
i.e., the Student-t test and the F-test. Whilst the former allows
comparison of means, the latter permits comparison of
variations.
Rollei Nikon
X Y 7 X Y Z
(cm) (cm) (cm) (cm) (cm) (cm)
RMSE 1,5 1,6 1,56 2,3 2,4 23
Standard Ls 1,6 1,5 2.3 2,4 2.3
Deviation
Maximum 3.1 2,9 3,5 4,3 4,1 4,0
Residual
Minimum -3,6 -3,4 -2,5 4,2 -4,0 -3,9
Residual
Table 4a. Statistical results by using PhotoModeler and the
digital cameras Rollei and Nikon
Inter!
—
Ta
I2 2 >. mm ez zl
Tabl
(MA
metri
phot
The
Niko
came
have
phot
the R
obtai
mon
Leic:
This
conv
The |
stere
obtai
To t
came
limit
2 to.
To |
meaı
form
archi
patri
patri
urba
Witk
phot
techi
stere
