gh a variety 
ıterest point 
ymation still 
acterized by 
ct modeling 
wittiness of 
ation for the 
this paper a 
. The whole 
JECTS 
ace model in 
ame implies, 
:d in an open 
> a change in 
  
lel 
set of points 
es of its form. 
or occlusion 
  
boundaries. In Figure 1, is clearly shown how breaklines are 
defined (heavy black color line) and how object's surface model 
is divided in homogeneous areas (shaded areas) bordered by the 
straight-line breaklines. 
2.2 The use of breaklines in close range problems 
There are important reasons why breaklines should be located 
in object reconstruction. The largest errors in the matching 
process occur due to breaklines inexistence, i.e. when a 
continuous object model is selected for use in surface 
reconstruction. Besides, if a continuous model is used without 
considering breaklines, the result of the image matching is not 
reliable especially at breakline locations. These reasons imply 
that in object reconstruction, breaklines should be detected first 
so that areas including breaklines should be modeled with a 
denser grid network or a triangulated network. A sparser grid 
network could be used for smooth areas where surface 
continuity is constant. These are the optimum definitions in 
order to minimize the number of unknown geometric 
parameters. 
Breaklines can be considered as the gross error of the 
mathematical model if surface of the object is modeled without 
considering breaklines, and smoothness constraints are used for 
reconstruction of the smooth object surface. 
By default, Photogrammetry considers the object reconstruction 
as one of the most fundamental procedure. There is no doubt 
that an increasing demand for rapid generation of surface 
models exists. Breaklines is a vital parameter for an accurate 
and reliable surface model reconstruction. Orthoimage 
production requires a known surface model, what is usually 
called as DTM, DSM or DEM. It is very important to initially 
locate the breaklines in automatic DTM collection because of 
the demand to have an approximate representation of the 
surface model. When image matching is used for automatic 
DTM generation, a large amount of gross errors in point 
location, occur due to the fact that a priori knowledge of 
breaklines position does not exist or is badly located. 
Moreover, when breaklines are located in the object space, the 
additional product of 3D wireframe model is created as well. 
This is useful information especially in cases of architectural 
issues where it can be used for image orientation even for a 
single image rectification. 
The above considerations indicate that it is vital in object 
surface reconstruction to detect breaklines firstly in order to 
have reliable and precise results. Besides, the additional product 
of the 3D object model that is created must not be ignored as 
well. 
3. THE ALGORITHMIC FRAMEWORK FOR 3D 
RECONSTRUCTION 
The developed algorithmic framework for 3D object 
reconstruction (Stylianidis, 2001) consists of 4 steps in turn. 
The “schematic flowchart” including all steps is presented in 
Figure 2. Image orientations (interior, relative and exterior) are 
not mentioned in the framework. They considered being as 
known parameters, in order to concentrate to 3D reconstruction 
section. 
  
  
  
  
  
Figure 2. “Schematic flowchart” o 
-217- 
  
2D model pu 
Image matching 
    
   
   
approximate model 
  
  
f the algorithmic framework