Mohammad Saadat Seresht
irallax extraction by Candy & Sobel composition method g) histogram of edge orientation testing of building
existence in region h) key-point extraction by Moravec method i.) line extraction and fitting.
tep is
10t be
ta and
built a
. This
| Figure 2: Outputs of test
7 CONCLUSIONS AND RECOMMENDATION
Coarse recognition of building regions and fine recognition of building details are two sequential steps
that offer a top-down idea for building recognition. This techniques results on improvement of reliability
. right and simplicity of work because if we know to exist only one building in selected region then much more
problems are obviated later.
Basic problems of coarse recognition of building regions are includes near or connected buildings,
Treated occluded areas and covered objects e.g. trees and recognition of dooryard of building that often hardly are
loes it. solved or are not solved.
oint in Because existing methods for building recognition and reconstruction are at beginning of way and still
are not solved completely, suggest that all aspects of it are studied for example multiple image techniques,
fusion of GIS data, using DSM for reconstruction, generic modeling by roof decomposition and working on
kill on more efficient semi-automatic methods.
ructed Non-parametric methods are recommended for building extraction because these techniques are very
trol is enable for complicate building that can not be modeled.
In future studies, it is better to concentrate on data fusion because today variety sources of data are
available.
The test results show that automatic coarse recognition of individual and simple building regions is
achievable. More complicated buildings, connected buildings and neighborhood features such as trees, may
rin the recognition. Since in Iran majority of buildings is connected, their recognition and reconstruction
meters may be the subject of further studies.
ymatic REFRENCES
ing 3D : orl.
) edge Baltsavias E., Mason S., Stallmann D., 1995. Use Of DTMs/DSMs and Orthoimages to Support Building
Extraction. Automatic Extraction of Man-Made Objects from Aerial and Space Images, p. 199-210.
e
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 797