In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010,1 APRS, Vol. XXXVIII, Part 7B 
247 
EOP : pitch [deg] EOP : roll [deg] EOP : ZL [meter] 
Figure 6. Platform modeling results, (a): X and Y, (b): Z, (c): 
roll, (d): pitch, and (e) yaw 
4.4 Georeferencing accuracy 
Since the image is monoscopic, the 3D ground restitution 
accuracy cannot be analyzed. Therefore, horizontal ground 
accuracy is analyzed by projecting image points to the ground 
with given ground height information. Each image point is 
projected to the ground using the estimated EOPs and the 
horizontal accuracy was computed. Table 3 shows the RMSE of 
the computed horizontal accuracy. The georeferencing result 
was accurate as the RMSE of ground coordinates is less than 1 
m, while Gauss-Markov modeling did not accurately estimate 
the perspective center positions and attitudes due to correlation 
between EOP parameters. 
RMSE 
X [m] 
Y [m] 
# of points 
0.47 
0.16 
186 
Table 3. Ground accuracy of the modeling 
5. CONCLUSION 
Airborne HSI systems have shown excellent performance in 
many commercial and defence applications. Most of the state- 
of-the-art systems are of the pushbroom camera type, and the 
georeferencing of these systems heavily relies on the GPS/INS 
based direct georeferencing. Under some rare circumstances, 
direct georeferencing may not be feasible, such as GPS denied 
situations. This study addresses these situations and proposes an 
image referenced georeferencing solution for pushbroom 
sensors. The approach is based on using high resolution satellite 
imagery, and the effort was motivated by the attractive 
properties of high-resolution satellite images, including high 
geolocation performance and image acquisition capability. The 
proposed method utilizes robust image matching, using a 
combination of the SURF and RANSAC techniques, and the 
platform modeling is based on the Gauss-Markov stochastic 
model. 
The SURF image matching performance with RANSAC was 
tested using simulated images and showed robustness by 
successfully mitigating low accuracy matching points. Next, the 
pushbroom sensor platform modeling was tested using 
simulated EOPs data and ground controls, which were obtained 
by simulated image matching. Test results indicated a good 
performance potential of the approach by showing high ground 
accuracy while EOPs could be estimated moderately.