142 
Benning, Effkemann, Schwermann 
As part of the outer work approximately 250 control points have been measured. Almost exclusively natural points on the building 
served as control points. These points were well distributed, so that the orientation of the images was made easier and the image 
blocks got a good fixation. Moreover, the selection of the control points always took into consideration a reliable recognition in 
black-and-white images. The documentation of the control points was done either directly in positives of special photographs or little 
sketches concerning the control points were explicitly made. Measuring and calculating the control points took up additional 4 days. 
The calculation of the coordinates was executed by KAFKA [Benning 1984], a high sophisticated program for adjustment tasks. Be 
cause the basic points were given in GauB-Kruger-coordinates, all control points were determined in the same coordinate system. As 
result of the point adjustment KAFKA achieved an average accuracy of 6 mm standard deviation in X, Y and Z. Regarding that al 
most all control points have been natural without signalization in most cases, this value seems to be really acceptable. 
3. PHOTOGRAMMETRIC OBJECT RECONSTRUCTION 
Two persons have been considered with the photogrammetric as-built documentation several years. Due to the complexity of the ob 
ject the measurement of the outer cathedral has been partitioned to 35 sub-projects. This partitioning simplified the daily work. Sin 
gle processes like image orientation and organizing the evaluation process are much easier, if the number of images and the amount 
of data are limited. Normally a project should contain less than 50 to 80 images. Regarding this recommendation the concrete sub- 
projects consisted of 30 - 50 images. 
As well known, before starting the graphical object reconstruction the orientation process has to be performed as a prerequisite. 
Therefore the control points are located in the images and additional tie points are determined. After measuring the image coordi 
nates of control and tie points the image orientation is done almost automatically in PHIDIAS (fig. 2). Firstly the program deter 
mines all initial values of the unknowns and, subsequently, the bundle adjustment takes place und delivers optimal values of the ori 
entation parameters. The adjustment process also yields statistical parameters related to accuracy, so that objective conclusions with 
respect to the measurement accuracy are possible. The global accuracy of the image measurements after bundle adjustment 
amounted to 15 to 25 pm with respect to a subproject. Sometimes this value was a little worse, but this result was accepted neverthe 
less, if the poor geometry of the image block did not allow an improvement. Image orientation needed 1 to 2 working days per sub- 
project on average, including additional tasks like aggregating of images, supplying of control points and transmitting of images 
from CD to hard disk. These tasks are time consuming and may not be neglected. Otherwise, image orientation in PHIDIAS is quite 
simple. Based on experiences an operator without special knowledge in photogrammetry is able to perform the orientation process 
absolutely independently after a short training period. 
Figure 2: The dialog box for photogrammetric image orientation in PHIDIAS 
Object reconstruction is the core work of the photogrammetric measurement, it needs the most time. Figure 3 gives an impression 
about the evaluation process with PHIDIAS. The digital images are loaded onto the screen and the evaluated design elements are 
displayed properly matching the image contours. After measuring distinct points in two or more images the threedimensional coordi 
nates of the object points are determined by forward intersection of homologous image rays. At the same time the CAD-model, 
which represents the object structure, grows step by step in shape of lines, arcs, circles, splines etc.. Due to the integration of 
PHIDIAS into the environment of a CAD-system like MicroStation, typical CAD related functions are immediately usable. This is a 
great help during object reconstruction, since measuring and constructing processes are simultaneously working hand in hand. Thus, 
the quality of the evaluation is essentially promoted. Moreover PHIDIAS comes with some special homogenizing tools. This means, 
that the raw evaluation data can be edited simply and effectively, for example: aligning linear elements in a vertical or horizontal 
position, restituting right angles between proper elements etc. The development of PHIDIAS has been permanently continued during