ÓÀ——MX 
Square 
lometrage 
(Area) 
A km? 
demon 
11.4 km? 
.2 km 2 
  
2.4 kin“ 
je detection 
ultitemporal 
teps. 
have been 
'opographic 
known map 
d study has 
iging to the 
and aerial 
een applied 
ed the land 
] accuracy 
ace through 
erent dates 
for studies 
eci 1994). 
sed in this 
is 
c———M—sÁ" 
ion /Scale 
1x10m 
1x30m 
2500 
5000 
rise 
  
The study reported here has been carried out by using 
the ERDAS V.7.5 in the laboratory of the Remote 
Sensing Division at the Istanbul Technical University. 
2,3. Methodology 
2.3.1. Preprocessing 
The procedure to rectify the image data sets to the 
universal Transverse Mercator (UTM) coordinate system 
involved the following steps. D Determination of Ground 
Control Points (GCPs) from 1/25000 scaled standard 
Topographic Maps and from the digital image data. © 
Computation of least-square solution for a first order 
polynomial equation required to register the image data 
sets to the UTM coordinate system © Resampling of the 
data sets using nearest neighbor algoritm.( Ehlers, 1990). 
The explanatory information related with the geometric 
correction in this research illustrated in Table 3. 
Table 3: Geometric Correction Information 
  
  
  
  
  
  
  
  
  
Sattelite Selected Used X Y Total 
Images GCP GCP RMS RMS RMS 
(m) (m) (m) 
LANDSAT 45 9 8.73 11.19 14.19 
TM 
  
2.3.2. Data Merging 
Landsat TM image data and SPOT P image data can be 
merged to effectively create enhanced multispectral 
images of high resolution. The resulting multiresolution 
images retained the spatial resolution of the 10m. SPOT 
Pancromatic referance of the Landsat multispectral data. 
Merge images can be shown in Figure 4. The enhanced 
detail, available from merged images has been found to 
be particularly important for visual Landuse 
interpretation. zu 
        
    
Figure 4: Merge images 
The aerial photographs covering the area have been 
obtained from the Istanbul Water Board Authorities 
(ISKI). The aerial photographs of the TEM motorway area 
located in the very near protected land are shown in 
Fig.5. 
Figure 5: Aerial photographs 
2.3.3. Classification 
In the classification procedure involved classifying the 
Landsat TM data sets using the convential supervised 
maximum likelihood classification algoritm. This study 
has been carried out with the seven classes level. These 
are water, forest, green area, bare soil, industry, road 
and urban. The training areas have been selected on the 
base of the information obtained from the field surveys, 
merge images, aerial photographs and existing maps 
and plans. For the signature evaluation, the mean and 
the standard deviation of every signature are used to 
represent the histogram in each Landsat Band of each 
potential class. By analyzing the ellipse graphs for all 
band pairs and then Landsat TM (3. and 4. Bands) 
selected for provide accurate classification results. The 
classified images are illustrated in Figure 6. The 
statistical results of the classified images for urban, 
roads, industry, forest, green area are classes according 
to the protected areas are tabulated in Table 4. Table 5 
gives the multitemporal land use statistics of the Elmali 
Water Basin and their percentages. 
259 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996