International Archives of the Photogrammetry, Remote Sensing and Spatia
(1
FAN
43
2$ ux
Maw 1503. 000000
Osa 123062324
Median 133 000001
Mes 19
Odaama Same: 77 762001
Stat Gama 450834
Dfdeos Sapwns 172113
Standart Say 5234
Figure.l Histogram of damaged and undamaged building.
The used photographs are taken in Gólcük in 1993 and 1999
and in Adapazari in 1994 and 1999.
The aerial photographs which are used have different scales.
The scale of the photographs before the earthquake is 1:35000,
after the earthquake is 1:18000. The general specifications of
both photos are given below:
e — Photographs of Gólcük; date 1993 and 1999 related
e Photographs of Adapazari, date 1994 and 1999 related
e The reports of the camera calibration belong to the all
photographs.
e Exterior orientation parameters belong to the all
photographs.
The critical point of this study is the vector data of buildings
before the earthquake are superimposed correctly to the aerial
photographs after the earthquake. For this reason, aerial
photographs after the earthquake are transformed into the geo-
coded coordinate system. In this stage, the model are formed
which belongs to the stereo photos after the earthquake.
Coordinates of photos and exterior orientation parameters
which have three dimensional coordinates belonging to the
vector data are shown in equation (1) and (2) as the parameters
of interior orientation are known, image coordinates are
obtained from the photograph coordinates by the application of
the affine transformation.
Basically, for an automatic recognition of building regions in
images, altimetric information in addition to planimetric
information have to be used because the usage of images only
can not completely lead us for building recognition (Seresht,
2000). Because of the earthquake the region is deformed. Also,
the elevations of the damaged buildings are changed. Between
the reality and theoretical model, undesirable differences have
occurred. In other words, after the earthquake photos, the vector
data does not superimpose correctly on the buildings. For that
reason, the user can modify regions interactively. The
deformation is the ground moved in different direction, but the
buildings are moved systematically in each different direction.
l Information Sciences, Vol XXXV, Part B7. Istanbul 2004
Therefore, to superimpose vector data and raster data that
belong to the buildings, the user can draw a selection polygon
and moved the selected area opposite of the deformation
direction.
a, (X —- Xo) + a, (Y - Yo) +ay(Z — Zo)
a CX — Xo) + a3, (Y — Yo) + aZ Zo) (1)
: a (X = Xo) + a„(Y-Yo)+ an (Z Zo)
= —C
ay (X — Xo) - a (Y = Yo) + ay, (Z = Zo) (2)
where c = focal length
X, Y, Z = ground coordinates
x, y = photograph coordinates
Xo, Yo, Zo = Photograph middle point
coordinates
a = orientation matrix
Image coordinates of the vector data belonging to a building
before and after the earthquake are calculated from the
photographs. Histograms which are related to these areas are
calculated automatically. Deciding whether a building is
damaged or not by using the developed algorithm, the
calculation of the number of pixels forming the building, grey
values average of a set of pixels, minimum and maximum pixel
grey values, mode, median average deviation and standard
deviation are calculated. These values are used for initial
parameters for the algorithm.
Figure 2. Histograms and photos of damaged and overlapped
buildings
When the region of Gólcük's histogram and statistical values
are analyzed, the differences of standard deviations before and
after the earthquake appear to be useful information in the stage
of decision. If the differences of standard deviations before and
after earthquake are used to determine the situation, the ratio of
accuracy is 9695.88 where the number of the building are 94
and 21 of them are damaged (Figure 2.).
640
Intei
In t
phot
dete:
used
appl
Espe
regic
builc
infor
the n
the a
If tv
orier
prob
algoi
In «
man:
earth
after
disas
Bayr
in €
Nati
Proc
Gling
Engi
Khar
2000
Sale,
Rem
183.
Kim
appre
vol. 1
Lami
comt
Poly?
Cons
Leto
Feat
Segn
ISPR
Raaf:
Datal
With
Sensi
Seres
Reco
of P|
XXX