Photogrammetric Recording 
In order to gain a desired result, objects should firstly 
be recorded either by metric or semi-metric cameras. 
In this case, photos can be scanned and stored digi- 
tally in a further step. On the other hand, digital CCD 
cameras are available today and may be used for this 
task, too. In addition, certain control information has 
to be acquired for the purpose of photogrammetric 
image orientation and restitution. 
Data Acquisition 
Distinct points and structure lines of the object are 
collected from stereo models using e.g. an analytical 
plotter, constituting primary data for surface modell- 
ing. Depending on the properties of the architecture 
concerned, various strategies can be applied to the 
data acquisition, e.g. computer-aided grid point meas- 
urement, profiling, single point registration, isoline 
tracing, automated or semi-automated line following. 
There are two things here to be taken into account. 
Firstly, adequate strategies should be used for parts of 
an object which are of different shapes. Secondly, an 
acquisition approach should be selected in accord- 
ance with the surface modelling method being con- 
cerned in further processing (Stephani/Tang, 1990). In 
case of digital images, image matching techniques may 
fully automate the procedure of point and line extrac- 
tion. 
Data Preparation 
Two tasks should be fulfilled in this procedure. First, 
measurement errors in primary data should be de- 
tected and eliminated. This can be done as well during 
the data acquisition using feasibility control utilities on 
the analytical plotter, as during the procedures of ob- 
ject modelling and visualization, which will be de- 
scribed in the following. Hence, a chained data flow is 
formed among these four procedures (cf. Fig. 1). Sec- 
ond, acquired data should accordingly be reformated 
and prepared for object modelling. 
Object Modelling 
In general, an object can be described by one or more 
DSMs (2.5D) or solid models (3D). For a detailed do- 
cumentation of architectures, DSMs are preferred. 
Surface modelling can be acomplished either by grid- 
ding or triangulating primary data (Stephani/Tang, 
1990). In contrast to some other applications, lines 
which represent the main structure of the architecture 
should completely be preserved in the resulting DSM. 
Therefore, triangulated irregular networks (TINs) are 
most suitable for surface description of architectural 
objects (Tang, 1989). The construction of a TIN of 
points and lines can be carried out either by a vector- 
   
ial constrained Delaunay triangulation (e.g. de Floria- 
ni/Puppo, 1988) or using the raster-based algorithm, 
which allows for a simple consideration of constraint 
edges and, as a matter of fact, improves the computa- 
tional complexion to a great extent (Tang, 1989, 1991, 
1992). 
Visualizati 
A DSM is an orderred set of data, which can be vis- 
ualized in different ways by means of computer 
graphics. Visualization can be realized either by vec- 
tor or by raster utilities. In the latter case, the pictorial 
aspect emphasizes the properties of the architecture 
being represented by the DSM geometrically (e.g. Ste- 
phani/Tang, 1990). The pictorial information may be 
carried either by digital images or by computer-gener- 
ated ones. Digital images (e.g. ortho images) reflect 
the natural appearance of the architecture, while pic- 
tures generated from the DSM give a realistic display 
of the model (e.g. shaded images) on the one hand 
and deliever a measure of the model (e.g. elevation 
layers, slopes and aspects) on the other hand. Combi- 
nations of different types of images may even enhance 
the realism of the display (Tang, 1991). 
Data Management and Analysis 
In order to meet requirements from different users, an 
architecture information system should be established 
(e.g. Rinaudo, 1988). It should consist of at least a vec- 
tor data base for primary data storage and a raster 
one for storage of images and DSMs. First of all, an 
efficient data base management should be available. 
Utilities for vector and raster processing should be in- 
cluded. In addition, appropriate user interfaces 
should be designed for purposes of analysis. 
3. EXAMPLES 
  
: d = : = 
Fig. 2. The DOME OF THE ROCK from south-west.