The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008 
840 
two alternatives provides similar results however, line-based 
scenario is more practical. 
Another simulation (S3) has been tested to check the object 
space accuracy if the navigation solution of the two data sets is 
good. Using the same photogrammetric configuration in SI, the 
image measurements are generated with appropriate noise level. 
The land data is then processed alone. Another result is 
obtained using both datasets. Table 4 presents the results in both 
cases and holds a comparison to present the percentage of 
improvement for the four common ground points T1 to T4. 
The results show that the van heading can be perfectly 
reconstructed. Both roll and pitch angles can be recovered with 
accuracy of 1’. The vertical direction accuracy is less than 1 cm 
and similar results were obtained for the direction perpendicular 
to the van trajectory. However, the state in the direction of van 
movement direction is not observable. In other words, moving 
the van in the forward direction will not affect the quality of fit 
and therefore the corresponding parameter will not be updated. 
This can be solved by having other lines which is not perfectly 
parallel to the van trajectory or having additional fewer point 
measurement either between the successive LMMS image set or 
connected with the airborne images if possible. 
Po. 
LMMS Only 
AMMS &LMMS 
Improvement 
2D 
3D 
2D 
3D 
2D 
H 
3D 
T1 
0.130 
0.131 
0.014 
0.016 
9.0 
0.9 
7.9 
T2 
0.163 
0.164 
0.037 
0.037 
4.4 
0.5 
4.4 
T3 
0.128 
0.129 
0.019 
0.021 
6.8 
0.8 
6.1 
T4 
0.034 
0.037 
0.016 
0.019 
2.1 
0.9 
1.9 
Ave. 
5.6 
0.8 
5.1 
Table 4: Object Space Accuracy Enhancement 
Plannimetric accuracy (Horizontal) is improved by the fusion 
algorithm with average of 5.6 times while the vertical accuracy 
deteriorates (compared to LMMS alone). This is not acceptable 
in terms of filtering and adjustment theories. This might be 
interpreted as a result of non proper weighting scheme or due to 
small number of points used to confirm the conclusion. In 
general the 3D accuracy of the common points is improved by 5 
times factor. The results of this simulation increase of LMMS 
operational range by fusion with high resolution airborne 
images, which has been limited by weak intersection geometry. 
The last performed simulation (S4) is used to support the 
innovative idea of using airborne images to improve/estimate 
the georeferencing of the land based systems when they exhibit 
large drifts due to GPS signal outages. A 500m trajectory was 
simulated. Two aerial images are simulated with two lines (two 
blue lines shown in Figure 5) in common with the land-based 
image set. These two lines simulating the lane line marking of 
the lane in which the LMMS van is moving. The linear features 
are back projected to the image space to generate line image 
measurements based on a true trajectory. Then, the navigation 
solution is contaminated with large biases (up to 20m) in 
position and several degrees in attitude. The data is then 
processed using the developed framework. The adjusted 
trajectory is compared to the true one. 
8. CONCLUSIONS 
The results presented in this paper reflects an ongoing research 
for establishing framework for integrating land-based and 
airborne mobile mapping system data-an integration scheme 
that received less attention in mapping, navigation, or 
photogrammetric literature. Both data sets are complementary 
in terms of information, resolution, and geometry. The 
developed framework adapts many of the existing tools to be 
generic enough to perform the proposed integration scheme. In 
this paper, we propose three vital applications for the fusion 
process. Firstly, land-based mobile mapping system (e.g. 
VISAT) can provide fast and efficient control points/lines for 
georeferencing airborne image sets. The presented georefencing 
strategy focuses on using linear features due to their extreme 
advantages for real life applications. Secondly, the proposed 
integration scheme enhances the 3D object space accuracy. 
Adding airborne to land-based data in one adjustment session 
reinforced the weak geometry in the imaging direction. This 
advantage potentially increases the operational range of land- 
based mobile mapping systems (usually was limited to 30- 
50ms). 
Thirdly, the framework proposes a practical strategy for 
improving land-based mobile mapping navigation accuracy in 
urban areas by fusion with airborne images. We focus on using 
the available linear feature like lane lines and road edges as 
matching entities. The obtained results are promising. More 
testing is needed to draw the guidelines for performing such 
strategy for bridging LMMS by fusion with airborne data. 
Our future works include the application object space constrains. 
Also, the earth fixed frame implementation will be tested for 
georefencing image sets which cover large areas. Optimal 
weighting strategy is included in our future plan. Once all the 
features of the proposed framework are implemented, the 
developed framework will be tested using real data. 
REFERENCES 
Cheng, W., Hassan, T. and El-Sheimy, N., 2007. Automatic 
Road Geometry Extraction System for Mobile Mapping, The 
5th International Symposium on Mobile Mapping Technology, 
Padua Italy 
Ebner, H., 1976. Self Calibrating Block Adjustment. 
Bildmessung und Luftbildwesen 44: 128-139. 
Figure 5: LMMS Bridging Simulation