19 
of 
maximum by compensation for the earth rotation. 
2. POLICY OF DEM GENERATION 
We have had the following policies on DEM generation. 
(1)DEM are generated operationally, if it is possible. 
(2)GCP aren't utilized with a basis. 
(3)Coarse DEM are utilized with a basis. 
(4)DEM of 30 m grid are generated. 
(5)DEM generation of Southeast Asia area is given high 
priority. 
By these policies, many relative DEM products will be 
possibly generated. But some problems are caused. 
(a)As ASTER is optical sensor, it is impossible to 
generate DEM in cloud area, water area, etc. Therefore, 
these area has to be detected automatically. 
(b)For assure to quality of product, abnormal DEM data 
that caused by mismatching have to be detected 
automatically. 
(c)The accuracy of elevation data depends on pointing 
knowledge of spacecraft because GCP isn’t utilized. 
Therefore, the performance of pointing knowledge had be 
clear. 
(d)Coarse DEM is desired one of high accuracy. As the 
best DEM product is DTED made by Defense Mapping 
Agency of USA, it isn't available. So, next better product 
is GLOBE product or DCW product. As the GLOBE 
product is 1Km Grid data. GLOBE will be selected as 
coarse DEM. 
3. The ALGORITHM 
The general flow chart shows in figure 1 and figure 2. 
This algorithm will be implement in ASTER Ground Data 
System. 
  
  
: ( Cloud Coverage Tabig (vNIRaN,3B Level1A Data ) 
Cloud&Water area 
Extraction 
  
  
  
  
  
  
  
  
(Cloud & Water Flag = DEM Generation 
EF | (Stereo Matching) [= 
DEM Browse | [“ApnormalValue] v 
Generation | Extraction | 
D | C (DEM Browse ) : (Abnornal Value Fg) | | 
| | I^. 7 
eem DEM X,Y,Z Set — 
  
   
   
   
    
  
    
   
  
      
  
   
    
  
    
Correlation V. 
Distance between : 
Coefficient 7: 
2 sight vector 
DEM X,Y,Z 
Data 
  
   
  
  
  
  
  
  
  
  
  
  
  
Figure 1 General Flow Chart (1) 
875 
  
(Dem X,Y. set) (Catalog 7» = (CGCP Data 
°° If there are GCPs 
  
    
  
    
   
Z data Conversion 
  
    
| Z data Conversion &| 
‘| Correction 
  
  
  
  
  
DEM Z Set 
  
  
Bias Corrected 
DEM Absolute DEM 
       
  
using a few GCP using some GCP 
South East Asia 
DEM DB 
| Specified products 
p p 
  
  
  
  
S _ 
Figure 2 General Flow Chart (2) 
We adopted the stereo matching method using area 
correlation technique. In a search of correspondence 
point, We utilized stereo matching method on three 
stages. Because it was a purpose that process time is 
saved and mismatching is reduced. So, First 
correspondence of tie point is examined using SSDA 
method. Second and Third correspondence is examined 
using correlation method. 
If GCP are available, Bias corrected DEM and absolute 
DEM are able to be generated. But GCP had to be 
prepared by not system administrator but user. Their DEM 
are treated as specific product. 
4. ASTER DEM PRODUCT 
Two kinds of DEM products will be distributed. 
(1)XYZ set product 
(2)Z set product 
XYZ set product is the Data Set that consisted of relative 
elevation data, correlation value and quality flags. 
The elevation is calculated by stereo matching, and the 
coordinate system is thing of earth fixation coordinate 
system. Therefore, it isn't grid data but random data. 
The quality flags are the index data which are indicated 
the area where elevation was abnormal, matching 
impossibility area. 
The correlation value is also one of quality index. If 
correlation value is low, probability that mismatching is 
occurred is high. 
XYZ set DEM will be generated 30sets per day. 
Z set product consists of only elevation data made from 
XYZ set DEM. 
It is grid data converted into map projection such as UTM, 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996