zraphs
nporal
:hosen
' north
sd as a
uercus
ferous
in the
area
PA
Ma
4 22
situe.
which
ds such
well as
'um(Sj)
is map,
e areas
ding to
ich tree
Musaoglu, Nebiye
DO 00-1 Qv tA A 02 D —
Figure 2. Test site
3 METHOD
In this study, Landsat TM images dated 12™ June 1984 and 18" July 1997 were used. First of all , boundaries of the
area of study were determined by applying geometric correction to the satellite images. Then, correction analyses were
made to the classification results by applying controlled and uncontrolled classification algorithms to the images and
temporal changes were examined.
3.1 Geometric Correction
Geometric errors that can be corrected using sensor characteristics and ephemeris data include scan skew, mirror-scan
velocity variance, panoramic distortion, platform velocity, and perspective geometry. Errors that can only be accounted
for by the use of ground control points (GCP) include the roll, pitch, and yaw of the platform and/or the altitude
variance (Bernstein, 1983).
In the first stage of the study, satellite images were transformed into Universal Transverse Mercator (UTM) co-ordinate
system by using 1/25000-scale standard topographic maps (Fig.3). In selecting the GCP’s to be used for
transformation, care was shown in homogeneous distribution of the sharp points and net differentiation on the map and
satellite image (roads, shores, etc.) and transformation equations were selected as a first-degree polynom. For the
geometric transformation, cubic convolution method was used. Co-ordinate transformation was done with + 0.5 pixel
root mean square (RMS).
(a) (b)
Figure 3: Satellite images of study area a) 1984 Landsat TM b) 1997 Landsat TM
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 941