In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
119 
Descriptions 
WorldView-1 
QuickBird 
Geoeye-1 
single 
CABA 
Improvement 
Geometric 
RFM 
RSM 
RFM 
WV-E 
0.87 
0.87 
0.00 
Model 
Level 
Basic 
Basic 
Standard 
WV-N 
0.64 
0.55 
0.09 
Date 
2007/11/25 
2005/12/23 
2009/9/20 
QBE 
0.50 
0.47 
0.04 
GSD (m) 
0.67 
0.63 
0.5 
QB-N 
0.62 
0.50 
0.11 
Image Size 
35170x 
27552 x 
19872x 
GE-E 
0.56 
0.51 
0.04 
23708 
29320 
16000 
GE-N 
0.43 
0.45 
-0.03 
No. of 
9/31 
9/20 
5/6 
KP-E 
0.94 
1.01 
-0.07 
GCPs/ICPs 
KP-N 
1.83 
2.03 
-0.20 
No. of 
16/24 
16/24 
16/21 
FS_1-E 
6.15 
2.65 
3.50 
TPs/ICTPs 
FS 1-N 
2.36 
1.28 
1.07 
FS 2-E 
6.92 
3.67 
3.25 
Table 1. Information Related to Test Data 
FS_2-N 
2.65 
2.15 
0.49 
Unit: m 
Descriptions 
Kompsat-2 
Formosat-2_l 
Formosat-2_2 
Geometric 
RFM 
RSM 
RSM 
Table 3. Absolute Accuracy Evaluation 
Model 
Level 
1A 
1A 
1A 
J.z Ceometrical Consistency 
Date 
2007/10/21 
2006/8/19 
2007/1/30 
The results fnr geometrical consistency between images are 
GSD (m) 
1.03 
2.01 
2.45 
. shown as Table 4. The CABA results for the WV, QB and GE 
Image Size 
15000x 
12000 x 
12000x 
images are approximately 0.5 m. The relative discrepancy can 
15500 
12000 
12000 
. be improved from about 3 m 
to less than 2 m for the KP image. 
No. of 2/12 9/18 9/22 The i m P rovement is significant for those two FS images. The 
GCPs/ICPs relative discrepancy for the former is about 9 m, with single 
No. of H/20 6/17 6/14 image adjustment. This decreases to approximately 3 m using 
TPs/ICTPs CABA. The second one is the same. The errors obtained are 
from 7 m to about 3 m. 
Table 2. Information Related to Test Data 
The validation include the absolute accuracy and the 
geometrical consistency between images. We use ICPs and 
ICTPs with the proposed model to evaluate the absolute 
accuracy and the geometrical consistency between images, 
respectively. The bias between the determined object 
coordinates and the true coordinates is calculated to find the 
root mean squared error (RMSE) for the ICPs. The ICTPs’ 
RMSE is obtained from the relative discrepancy for each image. 
To reveal the contributions of CABA, the results of the single 
image adjustment without collocation, herein named single, are 
also given. The Formosat-2, WorldView-1, Quickbird, Geoeye- 
1 and Kompsat-2 satellites images are labelled FS, WV, QB, 
GE and KP, respectively, in those results. 
3.1 Absolute Accuracy 
The results of absolute accuracy are shown as Table 3. 
According to Table 3, the absolute accuracy for WV is about 
0.9 m and better than 0.5 m for the QB and GE images. The 
absolute accuracy of KP is approximately 2 m. There is a small 
difference between the two methods for the higher resolution 
images. For the two FS images, the accuracy can improve to 
about 3.5 m when CABA is employed. 
single 
CABA 
Improvement 
WV-E 
0.56 
0.57 
-0.01 
WV-N 
0.80 
0.47 
0.33 
QBE 
0.42 
0.41 
0.01 
QB-N 
0.62 
0.57 
0.05 
GE-E 
0.49 
0.38 
0.12 
GE-N 
0.64 
0.51 
0.13 
KP-E 
1.14 
1.17 
-0.02 
KP-N 
2.82 
1.83 
1.00 
FS_1-E 
9.21 
2.65 
6.56 
FS 1-N 
6.52 
3.15 
3.37 
FS_2-E 
6.84 
3.22 
3.62 
FS_2-N 
2.34 
2.53 
-0.19 
Unit: m 
Table 4. Geometrical Consistency between Images 
4. CONCLUSIONS 
This paper combines DG and RFM for multi-sensor block 
adjustment. Two heterogeneous models with DEM as elevation 
control are integrated. The experimental results indicate that the 
proposed method can significantly improve the geometric 
accuracy as well as reduce discrepancies when multi-resolution 
images are used. Tests indicate that the proposed method should 
be feasible for real applications.