anbul 2004 
——M NE 
CH 
tegration of 
rom several 
] fused in 
ormation of 
We aimed to 
plication of 
id others. 
| à pixel to 
| as subtle 
sensor and 
lifferentiate 
attempted. 
sion results 
ructures. 
ove mobile 
very subtle 
geometric 
roperty (ax 
from DSM 
ts will be 
data were 
lop a novel 
lution, and 
e detection. 
JS40 stereo 
xad network 
1ethodology 
r fusion for 
oses. First 
ng was used 
n-made and 
images and 
» planned to 
and colour 
res. 
and textual 
both can be 
:hanges and 
he questions 
prove their 
ages. 
g individual 
> of ADS40 
spectral and 
c processing 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
(A ADGAD S C TABI ; ) 
  
Thermal HyperS | soe cal | nom now | 
GIS data 
Fous e = 
Tr 
| N Pixel 
pui * Matching 
| / Elevation / 
    
   
  
  
  
  
| Final Result £ Application 
Figure 1. Disparate high-resolution sensor data and GIS data 
automatic fusion for change-detection analysis. 
To refine the change-detection to fine level (subtle) other 
parameters such as NDVI obtained from hyperspectral images, 
Road data for moving objects location extraction and building 
data information were also used. After obtaining the final 
change-detection image a GIS analysis of attribute data was 
carried out. Later a pixel-by-pixel matching was carried out 
with TABI, and hyper-spectral images to know the 
landuse/landcover features. 
5. RESULTS 
5.1 Change-Detection by DSM Analysis 
Since DSM data plays a vital role in detecting subtle changes in 
landuse/landcover features, the obtained DSM data were 
matched for change detection purpose. DSM from stereo 
matching of ADS40 images of two dates was used at present. A 
broad change-detection image was generated first. By using 
DSM data obtained by pixel-pixel matching we tried to extract 
subtle changes in landcover for various applications such as 
automatic mapping and house taxing purpose. 
For the purpose of DSM extraction six stereo pairs of ADS 40 
images were used and DSM at 50 cm was obtained (figure 2. a, 
b) for two dates. Then by simple pixel difference estimation, 
changed areas were extracted (figure 2. c). The cyan colour 
shows areas where there is increase in elevation (structures 
demolished or no vegetation) and orange colour areas 
correspond to new structures or land cover features. 
Information, which can be read in change-difference, are the 
existence of the leaf, new-buildings, a removed building, and 
plant, a growth situation (after /before 2003 shedding leaves 
/2002 shedding leaves) and a moving objects. 
  
a). Year: 2002 
903 
  
b). Year: 2003 
  
C). Detested changes 
Cyan: Change but no features; Orange: New construction 
Figure 2. DSM obtained from pixel matching 
5.2 Thermal and Hyperspectral Fusion 
Figure 3 shows AISA and TABI data fused image. The AISA 
captured 66 ranges of bands and the PCA was fused with TABI 
thermal pixels. The temperature ranges from 33 deg. Cel (in 
orange) to 43 deg. C. (in Red). It can be observed from figure 3 
that the vegetated and cooler areas have lesser thermal values 
than the residential and industrial rooftops. 
The AISA sensor hyperspectral data was used to estimate the 
NDVI values (figure 4). Because of its dimensionality, 
hyperspectral data potentially provides the capability to 
discriminate between nearly any set of classes. 
  
Figure 3. Hyperspectral (PCA) and thermal data fusion result. 
Man -made structures show higher temperature values.