Figure 2: Photo capture
This configuration brings several advantages:
— the calculation of approximate values for bundle
ajustment and the check of blunders therein is easily
achieved
— the process of taking photos is simplified
— the definition of local coordinate systems for the
generation of the DOM can be done systematically
For this reasons the configuration looks as shown in fig.3.
Horizontal stereo models from eight directions have been
taken followed by corresponding photos with zenith
angles of 45° and 135°. To increase stability of the
ajustment two models with a zenith angle of 10° complete
the work.
top view front view
Figure 3: photo configuration
Image capture. The images were recorded with a Rollei
6006 metric camera with a 120 mm macro-lens and 11 x
11 réseau meshes (cf fig. 2). The mean exposure
distance of 86 cm results in an image scale.of about 1 :
8. The base length was chosen to 10 cm giving an
overlap for adjacent images of 80 %.
The images were recorded on a colour reversal film
(AGFA RS 120 Plus Professional) with a sensitivity of 50
ASA. In order to achieve a maximum sharpness a small
aperture (32) was used. The object was illuminated
72
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
simply by the ordinary room lights, because additional
spotlights would have produced specular reflections.
The model scale was introduced by a metallic ruler
positioned just in front of the busts plinth and as
information for the orientation of the model axis served
the edges of the plinth itself.
A/D conversion. Due to practical reasons, the anlog
images were digitized using two different equipments.
histogram
freq. T M
" ariadne's face
Figure 4: Gray value distribution
Using a Wehrli scanner the images were produced with a
resolution of 12.5 um, wheras the data from the Rollei
scanner came up with 15 pm. All scan processes were
optimized, in order to get a uniform distribution of the gray
values. Fig. 4 presents a typical distribution for a picture,
showing two parts corresponding to the objects surface
and the light background.
Photo triangulation. The triangulation was based on
manual coordinate meaurements performed at an
analytical plotter. Difficulties arised due to the lack of
texture combined with the considerable variations of the
perspectice for adjacent stereo models resulting in
special effort needed to identify individual object points. A
solution was found in the use of a digital CCD camera,
mounted in front of an eyepiece and storing the image of
the selected point together with the floating mark. For all
subsequent measurements at the same object point the
image was loaded onto a graphics screen allowing to
compare the visual impression in the actual image with
that one already given. This concept speeded up the
whole measurement procedure and reduced the amount
of misidentifications.
Over all 53 images 1090 point measurements were
registered. The triangulation was performed with the
program package BLUH coming up with an internal
precision of oo = 4.3 um for the image measurements
and an accuracy of oxyz = 0.15 mm for the
coordinates of the object points. The results show a good
geometric quality of the image configuration and are the
pre-requisite to establish a homogeneous DOM out of the
the locally evaluated surface parts.
GENERATION OF LOCAL OBJECT MODELS
Coordinate transformation. Although the object surface
will be expressed in a unique 3D-system (X,Y,Z), the use
of a stan
necessary
(XmiYmi-?
models (cf
=> standat
Z(X,Y)
proces:
= the me:
a long
inclinat
inclinat
object |
model |
The Z axi
chosen pe
that the lo:
of the obje
The use c
necessary
and point
individual |
m
dns [
Lm
Figure 5
Image m:
has been
based me
more or
described
Practical |
that the p
cases of
importanc
As alread
base to I
these val
difference
Consideri
surface p
parallaxe
expected
maximum
points, c
even on :
provide
convergel
magnitud
