2004 
—— 
This 
high 
Sof 
y the 
erent 
  
ation 
6 for 
erent 
in: it 
ntain 
tified 
mon 
tified 
on is 
and 
the 
good 
ectral 
same 
' data, 
| pan 
mage 
ginal 
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
[n order to control the geometrical quality of DOM, technical 
indices are needed as below: 
RMS of GCP: An important but never unique index to control 
geometric quality in the process of DOM production, it can be 
produced when rectification model calculates with GCP. RMS 
value varies with softwares even for same image and same 
GCP. 
Geometrical precision: An important index for DOM quality, 
itis checked with the high precision points from the field with 
GPS or other means. 
  
  
  
  
  
  
Type DOM precision 
IM/ETM Mae 
at* TOD 
SPOT? m In e 
SPOTS M La 4 s 15 m 
  
  
Table 4: Dom precision 
3.2.4 Image fusion: In order to get the best production 
enriched with geometric and spectral information, it’s very 
important to give a better combination for P and M images 
from many resolution images, and better fusion methods. 
Image combination: There are many kinds of combinations 
for fusion of P&M images, but it is suggested that the ratio of 
P&M image is better no greater than 4 times, for it's difficult to 
get a good geometric match for P&M and fusion results is not’ 
good if resolution ratio is too large. So the combination with * 
mark in table 5 is preferential: 
  
  
  
  
  
Dr 
um TM/ETM | SPOT2/4 | SPOTS 
TM/ETM + " 
SPOT2/ * 
SPOTS * 
  
  
  
  
  
Table 5. image combinations 
In this project, TM (M) + SPOT(P) and SPOTS (P+ XS) are the 
popular combinations. 
Image fusion: It is required that P&M images keep same 
resolution for fusion, generally Multi-spectral image is 
resample to the resolution of pan image. 
General fusion methods include Principal component analysis, 
HIS and Brovey, etc. these methods can be chosen according to 
the different image combinations or object characters. 
Figure 3-5 lists the fusion results with 3 fusion methods; it is 
obvious that HIS and Brovey methods are more effective for 
construction, 
  
277 
Figure 3. Brovey method Figure 4. HIS method 
  
Figure 5. Principal component analysis method 
3.3 Detection and extraction of changes 
Landuse changes can be derived in two ways: comparing 
satellite images in two phases (image-image) or satellite image 
and landuse map (image-map). It’s always a bottleneck to 
automatically detect and extract change no matter for 
image-image or image-map. 
This project will focus on studying methods for detecting 
change information; at the same time, it will also introduce 
some methods for the change extraction. 
3.3.1 Image-image: Spectrum aberrance method, Principal 
component difference method, subtract method, bands 
combination method and others can be used to show change for 
two-phase images. Among them, spectral aberrance method and 
subtract method are more popular because of they are easy to 
manipulate and the results are obvious. 
Spectrum aberrance method: Fuses pan image of second 
phase and multi-spectral image of first phase, change 
information can be remarkable with special spectral 
information. See figure6 as below (red arrow). 
  
Figure 6.Spectrum aberrance method Figure 7. Subtract method 
Subtract method: Using the second-phase image subtracts the 
first-phasc image with their gray value, white parcels in figure 
7 is the potential change information. 
But these two methods are only effective for new growing 
construction land that has obvious change in gray value. 
3.3.2 Image-map: For color landuse map with each kind of 
color represents one landuse type, it can be fused with image, 
the objects with abnormal spectrum maybe are change 
information (see the red arrow in figure 8):