Timur Murat Celikoyan
AUTOMATIC EXTRACTION OF OBJECTS AND THEIR INTEGRATION INTO AN INFORMATION
SYSTEM
T. Murat CELIKOYAN and M. Orhan ALTAN
Istanbul Technical University, Turkey
Faculty of Civil Engineering, Division of Photogrammetry
tikovant@sru. ins tu cducis . osltan@sty.ius. lu <à
Working Group III/3
KEY WORDS: Automation, GIS, Object Tracking
ABSTRACT
In photogrammetric practice, it takes a lot of time to measure continous objects manually. In order to gain time, many
objects like roads and buildings can be extracted automatically. To extract these kind of objects, many attempts are
done. The aim of this study is to develope a better algorithm for extraction.
The extraction process can be divided into two groups. The first one is to obtain the borders of a geometrically closed
object like a building, which can be defined as spatial objects, where the second one is to extract the unclosed objects
like roads etc. These two groups seem likely, but there are some differences about the algoritms, which are used to
obtain the borders of them.
As application, several images are worked on. The first two examples are images of damaged buildings in Dinar
Earthquake (1995) and Marmara Earthquake (1999), in which over 15.000 people were died and thousands of buildings
were destroyed.
For documentation and analysis of earthquake damages, an information systems is developed. As basic data, damaged
buildings are evaluated photogrammetrically. By these applications, damages on buildings are extracted automatically.
As further application, aerial images and remotely sensed data are worked on. Roads are extracted automatically on
aerial images. As a last example, coast line of Marmara Region is extracted from remotely sensed data.
1 INTRODUCTION
In recent years, many approaches and studies are done in order to extract and evaluate of objects form digital images.
Aims for these axtractions and evaluations were data acquisation for GIS as well as mapping. As basic data, aerial
images and remotely sensed data from multiple resolution and structure were used. In low resolution images, only the
road lines vegetation borders etc. can be extracted, whereby in high resolution images some profiles, road borders and
buildings.
For automatic road extraction Baumgartner et.al. (1999) has used aerial images from multiple scale and combined all
the results in order to obtain a road network. Wiedemann and Hinz (1999) has extracted roads from multispectral
satellite imagery. Further applications are made by Heipke and Straub (1999) in order to automatic updating a
Geographical Information System using automatic extraction methods from multispectral satellite imagery.
By automatic extraction, lots of approaches and applications can be given. What important is to finding the useful
algorithm for the purpose. In this paper, another algorithm for extracting objects from terrestrial and aerial images is
discussed and some examples are given.
2 ALGORITHM FOR EXTRACTING GEOMETRICALLY NON-CLOSED OBJECT
In the algorithm, pixel neighbourhoods are numbered as in Figure 1., in which having the same rotation in N, E, W and
S. Regarding the problems studied before, it has been exposed that every circle must have the same neighbourhood as
the neighbouring two circles. These neighbourhoods are chosen in the basic rotation as N, S, E, W.
164 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.