ing towards 
gulate strips 
ailroad line. 
or Mapping 
Ruekert and 
by Conrail 
TX). GPS 
il road by 
, which was 
5). The van 
e the Tail 
and analog 
ted by The 
1g a Kodak 
sensor) and 
ults of GPS 
coordinates) 
erate digital 
surrounding 
the aircraft 
3.5 m. The 
m. Image 
and 5 to 6 
measured 
We estimate 
han 1 pixel 
Using these 
interpolated 
"he variance 
s large. All 
each were 
  
tereo-vision 
  
The resulting product of this process is shown in figure 
6: the rail centerlines captured by the GPSVan are 
super-imposed on top of rectified and stitched aerial 
images. 
6. Conclusions and Recommendations 
A new technique for GPS controlled strip 
triangulation using linear features as control was 
presented. It was tested using both simulated and real 
data. The obtained results verify the feasibility of the 
proposed method. The method is based on the collection 
of ground coordinates along a linear object using a 
mobile mapping system, such as the Center for 
Mappings GPSVan. As mobile mapping systems can 
also be equipped with a stereo vision system for spatial 
feature collection from terrestrial image pairs, it is 
  
Fig. 6: Rectified and stitched digital aerial images overlaid by rail road center-line. 
possible to identify a few ground control points in each 
strip. This is of advantage, as it improves the accuracy 
of the results, especially if the interior orientation of the 
camera is not known precisely. Other applications for 
GPS controlled strip triangulation include the mapping 
of utility lines, pipelines, and highways. 
In the future this technique will be improved by 
automatically extracting the linear features in the 
digital, aerial images. In this case the whole 
triangulation process could be done without any human 
interference. Research is currently underway at the 
Department of Geodetic Science and Surveying to 
achieve this goal for aerial images of highways that have 
been surveyed with the GPSVan. Both control features 
as well as tie points must be identified automatically in 
this application, which will lead to the development of 
new feature matching and data association techniques. 
209