277 
Table 3 shows the average mean square error in X, Y of the 
orientation. It is clear that considerably accurate results were 
obtained for satellite image orientation, so the accuracy of 
image orientation has little effect on generating epipolar 
imagery. 
3.5 Results and analysis 
k The result of experiment 1 shows that no big differences 
exist between the coordinates obtained by the forward and 
inverse RPC models and extracted initially. This outcome 
indicates that the accuracy of forward and inverse RPC models 
is precision reached. 
2 Experiment 2 is carried out to generate the epipolar lines of 
satellite stereo images. It was found that the vertical parallaxes 
of the conjugate epipolar lines are not more than 1 pixel by the 
means of manual visual interpretation either of across-track 
stereo pairs or along-track stereo pairs. In which , along-track 
stereo pairs include the data from the same satellite sensors and 
different sensors. 
3 > Accuracy results presented and compared in Table (2) 
shows that the RMSEs in SPOT-HGR imagery and IKONOS 
imagery was 0.253 pixel and 0.027 pixel. This result is very 
good. But, the RMSEs in Radarsat imagery and SPOT-Radarsat 
imagery are comparatively large. 
4. CONCLUSIONS 
Finally, conclusions and recommendations for further research 
were presented. 
In the first experiment, the conjugate epipolar lines generated 
by the RPC model are highprecision. It conforms with accuracy 
requirement that use RFM model by way of sensor model. 
In the second experiment, for the five pairs of images, the 
vertical parallaxs of the conjugate epipolar lines are not more 
than 1 pixel by the means of manual visual interpretation as is 
shown from Fig(l) to Fig(7). But in the Accuracy Assessment 
Section, accuracy presented and compared in Table (2) shows 
that the RMSEs in SPOT-HGR imagery and IKONOS imagery 
are small, but in Radarsat imagery and SPOT-Radarsat image- 
pairs are comparatively large. For the radar imagery, it’s 
radiometric and geometric characteristics are special, and 
proper conjugate epipolar lines are inexistent. The author just 
to simulate its epipolar geometry in this method. The result 
shows that there is no means of generating epipolar imagers of 
radar by extended epiolar model. 
The epipolar resampling approach presented in this paper is 
based on the projection track method and using the RPC model 
as sensor model. This method is, therefore, appropriate for 
optical imagery such as SPOT and IKONOS imagery. In the 
practical experiments, it was shown that accuracy of epipolar 
generated better than half a pixel was achieved by the proposed 
approach. But for the SAR imagery, this method is not 
applicable. 
Desirable future work related to these methods includes: 
1. In this paper, the method of accuracy assessment is an 
interior accuracy; it is not strict enough to test the 
precision of entire epipolar line. The rigorous method is 
matching a point on one image to the other to obtain the 
conjugate point, and calculating the distance between the 
obtained point to the corresponding epipolar line. 
2. There is no need to resample the entire image in actual 
production, at present, Mapping software only resamples 
the area on the current Windows for the real time, so the 
local epipolar is the next step. 
3. The experiments in this paper do not establish the relation 
between the resampled imagery and initial imagery, that 
is to say, for the point on the resampled image, its 
location on the initial image is unbeknown. So this is also 
in the direction of future efforts. 
REFERENCES 
Cho, W., T. Schenk, and M. Madani, 1992. Resampling Digital 
Imagery to Epipolar Geometry [J], IAPRS International 
Archives of Photogrammetry and Remote Sensing, 29(B3): 
404-408. 
Danchao Gong, Yongsheng Zhang, 2003. Models and 
algorithms on processing of high-resolution satellite remote 
sensing stereo images. Acta Geodaetica Et Cartographic Sinica, 
4(83). 
Dowman, I., and J. Dolloff, 2000. An Evaluation of Rational 
Functions for Photogrammetric Restitutions [J], International 
Archives of Photogrammetry and Remote Sensing, 33 (Part B3): 
254-266. 
Fraser, C., and H. Hanley, 2003. Bias Compensation in 
Rational Functions for IKONOS Satellite Imagery [J], Journal 
of Photogrammetric Engineering & Remote Sensing, 69(1): 
53:57. 
Fraser, C., H. Hanley, and T. Yamakawa, 2002. Three- 
Dimensional Geopositioning Accuracy of IKONOS Imagery 
[J], Journal of Photogrammetric Record, 17(99): 465-479. 
Hong-Gyoo Sohn, Choung-Hwan Park,Hoon Chang, 2005. 
Rational function model-based image matching for digital 
elevation models. The Photogrammetric Record ,20,pp.366- 
383 . 
Grodecki, J., and G. Dial, 2003. Block Adjustment of High 
Resolution Satellite Images Described by Rational Polynomials 
[J], Journal of Photogrammetric Engineering & Remote 
Sensing, 69(l):59-68. 
Michel Fawzy Morgan,2004.5. Epipolar Resampling of Linear 
Array Scanner Scenes. 
http://www.geomatics.ucalgary.ca/links/GradTheses.html. 
Tetsu Ono,1999. Epipolar resampling of high resolution 
satellite imagery, http://citeseer.ist.psu.edu/270815.html. 
ACKNOWLEDGEMENTS 
During the research we gain lots of help from others. Thanks 
are due for the support from the National Basic Research 
Program of P. R. China (No:2006CB701302), the Natural 
Science Fund of P. R. China (No:40601084 and No:40523005), 
the Open Research Fund Program of State Key Laboratory of 
Satellite Ocean Environment Dynamics (No:SOED0602), the 
Open research subject of Key Laboratory of Geo-informatics of 
State Bureau of Surveying and Mapping, China International 
Science and Technology Cooperation Project: High-Resolution 
Stereo Mapping Satellite: Field Geometric Calibration and 
Application(No.: 2006DFA71570),the Open Research Fund 
Program of the Geometries and Applications Laboratory, 
Liaoning Technical University (2006004) .With those help, 
our research is able to go along propitious.