The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
10 
RMSE of resulting exterior orientation [mj 0 - RMSE of resulting exterior orientotion [m] 9 - RMSE of resulting exterior orientation 
Figure 9. RMSE on ground level as a function of ground 
distance to the camera 
5 out of 48 erroneous image points 
ground distance to control point [m] 
5. CONCLUSION 
Three different approaches for determining the exterior 
orientation of cameras for VIDS have been presented and 
thoroughly tested. The minimum space resection proved to be a 
very accurate and robust approach for determination of initial 
values and superior to the DLT. 
In general the exterior orientations derived from point based 
approaches result in a more then sufficient accuracy for a traffic 
monitoring sensor. The RMSE is less than 0.05m in object 
space up to a ground distance of 80m and less than 0.15m up to 
a ground distance of 140m. While the Newton method is unable 
to cope with erroneous control points, the Gauss Markov 
approach remains widely unaffected. Furthermore, using 
quaternions avoids possible ambiguities. 
Despite the encountered problems, using line features is a 
promising mean to determine the exterior orientation. Future 
research will focus on the stability and automation of 
calibration using line features. 
REFERENCES 
Abdel-Aziz, Y. and Karara, H.M, 1971. Direct linear 
transformation from comparator coordinates into object space 
coordinates in close range photogrammetry. ASP Symposium on 
Close-Range Photogrammetry, pp. 1-18. 
Anderson, B. and Moor, J., 1979. Optimal Filtering. Prentice- 
Hall, Inc., Enlewood Cliffs, New Jersey. 
Blackman, S.S., 1986. Multiple-target tracking with radar 
applications. MA: Artech House, Dedham. 
Brown, D.C., 1971. Close range camera calibration, 
Photogrammetric Engineering, 37, no. 8, pp. 855-866. 
Figure 10. RMSE on ground level as a function of ground 
distance to the camera with 5 erroneous image points 
Figure 11. Gauss Markov adjustment for points with a varying 
amount of control points 
Datta, T.K., Schattler, K. and Datta, S., 2000. Red light 
violations and crashes at urban intersections, Highway and 
Traffic Safety: Engineering, Evaluation, and Enforcement; 
Trucking and Motorcycles, 1734, pp. 52-58. 
Ernst, I., Hetscher, M., Zuev, S., Thiessenhusen, K.U., Ruhe, 
M., 2005. New approaches for real time traffic data acquisition 
with airborne systems, TRB Transportation Research Board, 
TRB 2005, pp. 69-73. 
Fischler, M.A. and Bolles, R.C., 1981. Random Sample 
Consensus: A Paradigm for Model Fitting with Applications to 
Image Analysis and Automated Cartogrphy, Communications of 
the ACM24 , vol. 6. 
Harlow, C. and Wang, Y., 2001. Automated accident detection 
system, Highway Safety: Modeling, Analysis, Management, 
Statistical Methods, and Crash Location, 1746, pp. 90-93. 
Kastrinaki, V., Zervakis, M. and Kalaitzakis, K., 2003. A 
survey of video processing techniques for traffic applications, 
Image and Vision Computing, vol. 21(4), pp. 359-381. 
Klein, L.A., Kelley M.R. and Mills, M.K., 1997. Evaluation of 
overhead and in-ground vehicle detector technologies for traffic 
flow measurement. Journal of Testing and Evaluation, 25(2): p. 
205-214.