3. Istanbul 2004 
' combined into 
e combined into 
mbined into one 
1ed into another 
nd image group 
respectively. 
re necessary for 
. If we set four 
cent orbits, then 
lock. 
ages, the initial 
points can be 
jn parameters of 
elow step: 
the orbit into a 
ntrol points to 
not include the 
ned images. 
yigger combined 
tial value of the 
block. 
dinates of all tie 
meters of each 
DJUSTMENT 
three orbits. The 
1). The block has 
jin the test block 
UTM coordinate 
600 meter in the 
geometric model 
and 8 GCPs three 
the Table 1. The 
model (the forth 
the purpose of 
KONOS imager 
:k points (m) 
   
  
  
  
  
  
My Mz 
0.58 0.87 
0.72 1.15 
0.85 0.93 
  
  
  
From the adjustment results listed in the Table 1, one can draw 
the conclusion: 
I. The accuracy of the block adjustment with strict 
geometric model can reach to the level corresponding to 
the ground resolution of the image and the accuracy can 
be improved by increasing the number of the control 
points. 
2. The planimetric accuracy of the block adjustment with 
strict geometric model is better than the accuracy of the 
adjustment based on RPC model. 
3. The height accuracy of the block adjustment with strict 
geometric model is coequal to the accuracy of the block 
adjustment based on RPC model. 
42 TEST RESULTS OF BLOCK WITH OTHER HRSI 
The second test block has six HRSI images in one orbit. The 
pixel size of these images is three meter and the photograph 
area is about 2700 km^ (90*30 km). As shown in the Figure 4, 
there are 39 GCPs in the block. All the GCPs are in the UTM 
coordinate systems. The terrain height variation is about 300 
meter in the block. 
  
  
je 9n km >| 
5801 / 5802 5791 f 5792 3781 / 5782 
v o 
3 o o 
e o o o 
> a e 5 o o 
9 9 2 e 
e e a 
d = 
o 
9 5 al © 9 o 
o " o 
o 5 
D o o 
v o o 
  
  
  
  
  
  
  
  
  
  
  
  
o o 
> 
e 
e 
© 
o 
» 
o ? 
(b) 10 GCPs 
o 
o 
o 
€ 
o 
* 
  
  
  
  
  
  
  
(c) 6 GCPs 
7 
Figure 4: The location of GCPs with the third test block. 
The block has 6 images in two orbits. There are 39 GCPs in 
all. The block adjustment has been tested for all the 39 
The block adjustment has been tested for all the 39 GCPs, 10 
GCPs and 6 GCPs three cases respectively and the test results 
are listed in the Table 2. From the result we can see that the 
planimetric accuracy of the block adjustment reach the level 
corresponding to the ground resolution of the image. But the 
accuracy of the height is a bit worse. One of the reason is the 
image is not very clear (Figure §), The other reason is the poor 
83 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
location of the GCPs: there are no GCP in the center of the 
block while the span of the block is 90 km. 
Table 2: The block adjustment results of the CBS-2 imagery 
  
  
  
9, Control points (m) Check points (m) 
Pixel i ; 
1 N Mx My Mz N Mx My Mz 
0.24 39 3.14 2.28 6.74 
  
0.22 10 1.46 0.88 4.22 29 3.79 2.93 7.04 
  
0.23 6 1.48 1.09 2.61 33 5.98 3.79 6.44 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Figure 5: The image windows (zoom ratio is 3) in the 
second test block. 
Table 3 shows the block adjustment results of the single models 
(model 5781/5782 and 5801/5802) in the third block. In each 
model, all the control points are used to calculate the orientation 
parameters. 
Table 3: The block adjustment results of the single model in the 
third test block 
  
  
  
  
  
  
  
Control points (m) 
Model 
N Mx My Mz 
5781/5782 19 3.57 2.19 4.19 
5801/5802 20 4.87 2.34 7.92 
  
  
  
  
Compared to the results shown in Table 2, we can see that the 
accuracy of the block adjustment with multi models reach the 
same level of the accuracy of the single model. 
5. CONCLUSION 
In this paper, a new approach of the block adjustment based on 
the new strict geometric model for HRSI is introduced. In order 
to reduce the number of necessary control points for the 
calculation of the initial value of the block adjustment, a 
procedure with three steps is proposed. The new approach of 
the block adjustment has been tested for IKONOS and CBS-2 
images respectively. All the test results show that the accuracy 
of the block adjustment can reach to the level corresponding to 
the ground resolution of the image. 
Compared with the block adjustment based on RPC model, 
though the block adjustment based on RPC model may need 
fewer control points, this new approach has larger potential in 
the application of HRSI because most of high-resolution 
satellite imagery has no RPC parameters supported.