International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B6. Istanbul 2004
parameters that can be controlled during this process. The main
parameter is step of regular grid over which the model is
created. The detail of processed model is another variable
parameter. It corresponds to the maximum difference in the
heights between two adjacent points in the model. The
appropriate selection of this parameter is very important for the
stability of generation and the creation of an adequate model.
Model Editing is very useful feature of the automatic DEM
creation procedure. In this process it is possible to arise a fail of
correlation function and absence of terrain data for some parts
of the model. Another type of error is the generation of wrong
values. The raster matrix of correlation values at separate points
is used for estimation the reliability of the model. Analysing
this matrix it is possible to find areas of missing or wrong
values and to edit them. Different algorithms for DEM
restoration are possible. Some of them are based on filling with
fixed values and another on interpolation from boundaries of
wrong areas to the inside points. Training for proper editing of
model is important for improvement the quality of the DEM.
Georeferencing (geocoding) of model is necessary when DEM
is produced from stereo images and created model is in a local
co-ordinate system. If the georeferencing step is omitted then
for every pixel of the transformed image it is necessary to
calculate coordinates in the model co-ordinate system that
increases the time of processing.
2.4 Orthophoto Production
Orthorectification requires knowledge of external orientation
elements of digital photo parameters of projection and height
data from DEM. The method of interpolation to calculate output
pixel value can be selected. The visual quality and measurement
characteristics of transformed image depend on it. The speed of
processing depends on the order of interpolation and number of
input pixels involved in interpolation process. The practical
training for different types of transformation is possible. The
mosaicking of images creates a full image from overlapping
orthoimages, which are transformed in the same target co-
ordinate system. In the process of mosaicking it is possible to
select the method for choice of cutting line — operator
controlled or manual. Another feature that could be controlled
is the method of equalizing of radiometric characteristic of
images based on transformation of histograms in the
overlapping zone of two images. The results of image merging
can be compared depending on different methods of image
equalizing.
2.5 3-D digitising
The process of stereo digitising in digital images is very
important in process of discrete data capturing. This process can
be done in initial images or in the epipolar images. For this
process there are important possibilities for creation of different
graphic objects — lines, arcs, closed polygons, polygons with
rectangular sides. Another group of features are space editing,
snap in 3-D space. The flexibility of digitising system depends
on the attributes of the graphic objects as layer, colour and
width. Training of students must be done for different modes of
editing and for different types of primitive objects.
3. TRAINING IN DIGITAL CLOSE RANGE
PHOTOGRAMMETRY
The technological process of application the digital
photogrammetry for documentation and photo realistic 3-D
modelling includes the main steps of photogrammetric
48
processing. The specific procedures are connected with
enhancement and radiometric correction of images. These
processes are traditional in processing of images in remote
sensing data acquisition and processing.
The sequence of processing and applied procedures is described
in (Marinov, B.D., Hristova, G.E., 2001). It includes following
steps:
Photogrammetric shooting and survey measurements.
Creation of a grid model.
Processing the digital images of separate surfaces (generation
of photo-textures).
4. Association of the grid model with the photo-textures.
|;
2.
3.
Depending on the used software products and initial data the
stages of grid model generation and assigning of images to the
surface to be connected. In case of separate generation of grid
model and photo-textures it is possible to use general - purpose
software for digital processing of colour images of surfaces.
3.1 Processing of digital images
Digital processing of images includes colour balance and grey
level value correction of images. For these purposes was used
Photoshop. Visual comparison of adjacent images was made.
The curves for transformation are adjusted based on analyses of
colour histograms of merged images.
The Intergraph program Advanced Imager was used for
merging of images. This is a general purposes image processing
software. Functions for image rectification and transformation
are used in conjunction with possibilities for transformation of
image - to - image and merging of images. Training process
includes the appropriate selection of tie points between images
and usage the different order of two - dimensional
transformation function. Final results of this processing step are
mosaics of images, which are corresponding to separate
surfaces. The transformation of image of surface is converted to
grid model by usage of specific features on the surfaces. In
cases of documenting the preserved architectural and historical
monuments when signalising of tie and control points is not
allowed the natural features are selected. They are chosen in
such way that they could be easy distinguished in digital images
and in the object. These features are mapped in process of
creation of grid model as vector objects.
Rectification of surface photo-textures is made by function for
image to map transformation of MGE Advanced Imager
(Intergraph, 1995b).
Transformation of photo-textures of curved surfaces requires
special procedure. They are suggested different approaches for
transformation of images over the curved surfaces. For
rectification of images over the cylindrical surface of vault of
the church entrance is used two stage procedure (Marinov B.D.,
2000). The first step is the rectification of separate images to
the unfolded grid model of the arc. The second step is merging
of rectified images in one common photo-texture and
transformation of the whole image to the grid model. It is fit
over the curved surface of grid model by program for space
modelling. Experiments with usage of high order 2-D or spline
transformation of merged images were not successful. The
reason is that it is not possible always to find enough number of
suitable located control points for such transformation.
