REESE 
   
LEO miS 
— 
  
  
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
  
  
  
  
  
  
  
Nominal collection 41.2363 10.5023 
azimuth (deg) 
Nominal collection 69.6502 63.2446 
elevation (deg) 
Sun angle azimuth (deg) 138.2219 166.2923 
Sun angle elevation (deg) 67.2403 41.5399 
Nadir angle (deg) 20.3498 26.7554 
Image size (pixels in row, 11,004x11,0 | 11,004x11,0 
column) 00 00 
Reference height (m) 206.78 208.04 
  
  
  
  
  
While the scene named as mage I was acquired on July 2002, 
[mage II was taken on October, 2002. These images are almost 
covering the same area on the ground and studied part of the 
Image II is shown in Fig. 1. In the upper part of the Ikonos 
image, Black Sea is lying and other parts of the image includes 
central part of the Zonguldak city which covers nearly 
10x10km area with the elevation range up to 450m. When the 
images first received, they were analysed for selecting suitable 
GCPs distributed on them uniformly. As a result of this 
determination, 43 distinct GCPs were measured by GPS survey 
with an accuracy of about 3cm. Since those points can be seen 
very well on the images, they were selected as building corners, 
crossings, etc. Because of the fine resolution of Ikonos imagery, 
many cultural features can be identified and used as GCPs. The 
manual measurements of GCPs' image coordinates were carried 
out by GCP Collection Tool under PCI Geomatica-OrthoEngine 
software package with zoom factor 4. Thus, accuracy of image 
coordinates could be expected in the range of 0.2-0.3 of a pixel. 
Geometric correction of these by different mathematical models 
produced the rmse values of about | pixel. 
Results of geometric correction of Ikonos Geo-product imagery 
has been given in detail in Buyuksalih, et al., 2003. 
  
Figure 1. Ikonos pan-sharpened image of the study of area 
Before analysing the Ikonos image with eCognition it was 
enhanced by applying a pan-sharpening method used in PCI 
system. This method makes it possible to benefit from the 
sensors spectral capabilities simultaneously with its high spatial 
resolution. Thereby the first principal component of the four 
spectral IKONOS channels (4m resolution) was substituted by 
the Im resolution IKONOS panchromatic channel. This new 
combination of principal components then was re-transformed 
applying an inverse principal components transformation. 
3. IMAGE SEGMENTATION AND CLASSIFICATION 
BY ECOGNITION V3.0 
Segmentation is the main process in the eCognition software 
and its aim is to create meaningful objects. This means that the 
shape of each object in question should ideally be represented 
by an according image object. This shape combined with 
further derivative colour and texture properties can be used to 
initially classify the image by classifying the generated image 
objects. Thereby the classes are organised within a class 
hierarchy. Each class can have a sub- or super-class and thus 
inherit its properties from one or more super-classes or to its 
subclass (es). With respect to the multi-scale behaviour of the 
objects to detect, a number of small objects can be aggregated 
to form larger objects constructing a semantic hierarchy. 
Likewise, a large object can be split into a number of smaller 
objects which basically leads to two main approaches of image 
analysis: A top-down and a bottom-up approach (see Benz, U., 
et al., 2003 and eCognition User Guide, 2003). 
In eCognition both approaches can be realised performing the 
following steps: 
e Creating a hierarchical network of image objects using the 
multi-resolution segmentation. The upper-level image 
segments represent small-scale objects while the lower- 
level segments represent large-scale objects. 
e Classifying the derived objects by their physical 
properties. This also means that the class names and the 
class hierarchy are representative with respect to two 
aspects: the mapped real-world and the image objects 
physically measurable .attributes. Using inheritance 
mechanisms accelerates the classification task while 
making it more transparent at the same time. 
e Describing the (semantic) relationships of the network's 
objects in terms of neighbourhood relationships or being a 
sub- or super-object. This usually leads to an improvement 
of the physical classification res. the class hierarchy. 
e Aggregating the classified objects to semantic groups 
which can be used further for a so called ‘classification- 
based’ segmentation. The derived contiguous segments 
then can be exported and used in GIS. The semantic 
groups can also be used for further neighbourhood 
analyses. 
These steps describe the usual proceeding when working with 
eCognition. While the first two steps are a mandatory, the latter 
two steps may be advisable according to the user's objectives 
and content of the image. 
In the segmentation phase, following parameters should be 
assigned as accurate as possible, of course, suiting with the 
reality: 
     
    
     
    
   
     
    
   
   
   
    
   
   
   
   
   
    
   
   
    
   
  
   
   
   
    
   
   
   
  
   
  
  
  
  
    
   
  
    
    
        
        
Internati 
ci 
° Sca 
ave 
ma) 
the 
e Col 
VS. 
adj 
hor 
e Sm 
larg 
sha 
Segmen 
eCogini! 
neighbo 
as class 
describe 
which tl 
describe 
do belor 
Thereby 
to desci 
feature : 
describe 
of fuzz 
combin: 
reflect t 
level c 
generat 
simply 
level. C 
property 
clearly : 
Object- 
parame 
scale in 
the sut 
segmen