ems em-
covision
h the co-
d on the
>s of cor-
e usually
measure-
| simulta-
ited with
lis meas-
nal capa-
methods.
late
(LSTM,
position
image of
s the po-
t squares
ind patch
approxi-
1 be indi-
lode. In
1 the data
transfor-
7 shows
he patch
ce of the
g LSTM
lications
| spacing
'curacies
letry one
of an ar-
| the ver-
e shown
t locates
and then
do
object is
process,
Beside a
| surface
a. tracked lines.
Figure 8: Example for feature location via line tracking.
model of the object is created without any additional in-
terference in the measurement process.
Figure 8 shows an example of feature location via line
tracking. The approximate position of the feature is indi-
cated by the operator in the shape of a polygon with the
cursor. Starting from that positions, the linear boundaries
of the feature are tracked by an algorithm using the first
partial derivatives of the image (see figure 8.a.). A
straight line is fitted onto each segment of the polygon
(see figure 8.b.). The coordinates of the vertices are com-
puted automatically as intersections of the appropriate
straight lines (see figure 8.c.). Thereafter the geometric
and semantic information of the feature is known. The
operator needs only to judge the scene qualitatively while
indicating the initial values for the algorithm. In principle
this could be provided by a suitable CAAD-model of the
object. The measurement itself is performed by the com-
puter. This measurement technique delivers image coor-
dinates for the vertices of architectural features with a
precision of 1/ 10% to 1/20 of the pixel spacing:
6. PHOTOGRAMMETRIC ANALYSIS AND
RESULTS
The photogrammetric analysis of the analogue and digital
image data was performed with different hard- and soft-
ware. The Digital Photogrammetric Station was used for
processing the digital images and the analytical plotter
Wild AC3 was used for measuring of the analogue imag-
es. The image data of the Canon CI-10 and the analogue
images of the Rollei 6006 were used to determine the ge-
ometric and semantic information of the church. The Rol-
lei images of the front facade were digitized to compare
these results with the results of the analogue technique.
The reference data for the photogrammetric network was
determined by theodolite. Therefore 20 signalized points
were fixed on the facades and used as control points.
‘Each facade with the ideal camera arrangement (see fig-
ure 3) includes four signalized points and each facade
With the non-ideal arrangement has six signalized points.
The precision of the geodetic reference coordinates is
1.2 mm in plane and 4.0 mm in height.
b. extracted straightlines of feature. — c. extracted vertices of feature.
6.1. Calibration
To obtain precise calibration parameters for the cameras a
testfield calibration was performed beforehand. For the
Canon CI-10 ten additional parameters were determined
using 35 images and 162 object points. The additional pa-
rameters are the three parameters of the interior orienta-
tion, a scale factor in x-direction, a shear, and parameters
for the radial and decentering distortion. The relative ac-
curacy of this testfield calibration is 1 : 10'000. The Rol-
lei 6006 was calibrated using 8 images and 58 object
points. The relative accuracy of this calibration is
1: 77000. The higher relative accuracy attained with the
CI-10 is attributable to the high redundancy of 26 rays
per object point. The results of the testfield calibration
were then used for the photogrammetric analysis of the
church.
6.2. Measurement
Within DIPS the image coordinates of the signalized
points were determined by point location with LSTM
(chapter 5.2.). The coordinates of architectural features
were determined by feature location via line tracking
(chapter 5.3.). In some cases, due to the insufficient reso-
lution of the images taken with the Canon CI-10, these
measurement techniques could not be used. Then it was
necessary to measure these coordinates by manual point
location (chapter 5.1.).
The analogue images were measured with the analytical
plotter Wild AC3. All relevant colour slides which pic-
tured the facades of the church have been processed on
the stages of the AC3 for stereo measurements without
exchanging the slides. The signalized points were used to
determine the orientation of the stereo pair. The acquired
data was displayed on a workstation, which is connected
to the AC3. Moreover, the operator was able to edit the
current measurements at any stage of the 3D-interpreta-
tion. For architectural processing the photogrammetric
results can be transformed into a standard format and
transferred to AutoCAD.
