tal 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B6. Istanbul 2004 
  
Depending on their power the systems in digital 
photogrammetry could be divided in three groups: low level 
systems, intermediate level systems and high-level systems. 
The low level systems usually do not have high automation of 
measurement and support only the monoscopic or anaglyph 
methods of measurement. The DOS version of Digital Video 
Plotter (DVP) is a system of such type. The intermediate set of 
digital photogrammetric systems is most numerous. It includes 
the systems: VirtuoZo of Supresoft, SISCAM, DVP (Windows) 
of DVP Geomatics, OrthoEngine of PCI Geomatics, Photomod 
of Racurce. They are usually universal systems that support 
othophoto production technology and digitising functionality. 
They utilise monoscopic or stereoscopic measurements of tie 
points and control points and the stereoscopic measurements in 
digitising. Almost all systems allow producing DEM from 
stereo but the diversity of editing of DEM and reliability of 
produced DEM is different. The supported function of 3-D 
digitising includes editing of the discrete data and the flexibility 
of attribute generated data. The high level systems usually have 
all possibilities of system of intermediate class but also possess 
high level of automation of separate processing and in their 
operation there are included some features of artificial 
intelligence to improve the accuracy and reliability of obtained 
results. The Zlmaging of ZEISS-Intergraph and Digital 
Photogrammetric Workstation of Helava belong to this class of 
systems. 
2. STEPS OF PROCESSING 
Different steps of processing must be developed to achieve the 
final photogrammetric product. There are two main 
technologies that are implemented in the digital form — 3-D 
digitising and othophoto processing. They correspond to the 
two main traditional photogrammetric technologies. Both 
technologies require determination of the elements of the outer 
orientation of the images (digital photos) in the object 
coordinate system. À specific step of the processing is the 
preparation of epipolar images that are more convenient for 
extraction of Digital Elevation Model from stereo couple and 
for more accurate 3-D digitising. 
2.1 Digital Triangulation 
The process of triangulation in digital systems follows 
traditional steps specific for analytical photogrammetry. 
Depending on the used model is possible to separate inner 
orientation and absolute orientation of photos in programs 
based on the bundle block adjustment, or inner, relative and 
outer orientations in programs based on the method of models. 
The measurement of tie points and control points is applied in 
all cases. Inner orientation requires measurement of image 
coordinates of fiducial marks and appropriate choice of 
transformation methods. Depending on the type of the camera 
they are central, angle or reseau. Training for improvement the 
measurements of fiducials marks based on residuals is 
important feature of digital system. The applied transformation 
method allows usage of different number of parameters. The 
comparison between orthogonal, affine and perspective 
transformation is advantage but it is not implemented in all 
systems for digital photogrammetry. In some systems for digital 
photogrammetry are implemented algorithms for automatic 
measurement of fiducial marks. They are based on image 
matching algorithms between images of fiducial marks and 
their patterns. The measurement of tie points could be done 
manually or automatically. The automatic measurement of 
corresponding points for relative orientation is based on the 
4] 
image matching. The training for this process includes proper 
selection of tie points position and accurate measurement of 
corresponding points in stereo images. Estimation of the errors 
in relative orientation can be done by residuals in vertical 
parallax or more precisely by variance-covariance matrix. The 
measurements of control points are necessary for outer 
orientation of photogrammetric model. The knowledge about 
convenient choice of their position is important for good quality 
of the training process. The automatic identification of control 
points is more difficult. Usually the operator makes the 
measurements. The control points’ residuals are criteria for 
proper selection and positioning on them. The monoscopic or 
stereoscopic measurements of tie points and control points are 
possible. If measured point occurs only in two photos then 
stereoscopic measurement or automatic matching is preferable. 
The quality of stereo viewing systems is important for visual 
stereo measurement. If point has three and more images in 
different photos monoscopic measurement and simultaneous 
visualization of all points gives better results. For automatic 
measurement correlation matching in different direction is 
preferable. Except in strip overlap the side overlap between 
strips is possible to be used. The accuracy of measurement can 
be controlled by recalculation of adjustment after each change 
of image co-ordinates. 
2.2 Epipolar transformation 
Generation of epipolar images is very important stage in digital 
photogrammetry. In analytical photogrammetry measurements 
are made over the images that are not optically oriented. Only 
in some models of Analytical plotters are included subsystems 
for rotation of images by Dove prisms after measurements of 
firsts two tie points (Kraus, K., 1993). The relative orientation 
in analytical photogrammetry is applied only in mathematical 
model. In digital photogrammetry the transformation is applied 
to the whole images not only to the measured points. These 
techniques allows to process corresponding points in the 
parallel image rows of transformed images. This procedure is 
based on the theory of epipolar geometry. The epipolar 
transformation is necessary stage not only for DEM extraction 
and orthophoto transformation, but for 3-D digitising too. The 
process of stereo visualisation and measurement is easily to be 
made over the epipolar images rather than in the initial ones. If 
images are not epipolar transformed then it is necessary to 
adjust the relative vertical position of two stereo images 
depending on the position of measured point in the stereo 
model. Such technique is not implemented in all system for 
digital photogrammetry but it is obligatory for system 
producing digital orthophoto by usage of DEM created from 
stereo couple. 
2.3 DEM Creation 
Digital elevation models that are used in Digital 
photogrammetry could be produced from different sources: 
survey data, vector topographic maps, analogue instruments 
with digitising equipment, analytical photogrammetric systems 
or digital processing of stereo photographs. In the most cases a 
raster terrain model with different resolution is generated. The 
break lines of the terrain are measured to achieve more 
adequate terrain model. The utilisation of structural terrain lines 
is not necessary in cases of very high resolution of raster terrain 
model. Such models are generated if DEM is produced from 
stereo photographs. Extraction of DEM from stereo images is 
based on the correlation matching of corresponding points in 
rows of epipolar transformed images. There are several