Sergey Zheltov 
  
  
  
(b) 
Figure 7. Examples of detection. (a) A car tire. (b) A pedestrian. 
7. CONCLUSION 
Method of using the difference in orthophoto images is well known for refinement of Digital Terrain Models [F. Raye 
Norvell, 1996]. However, this paper shows that similar idea can be used for object detection in a scene. 
Our method has been tested on real data under different conditions. The problem with contrast objects belonging to the 
underlying surface (for example, drawn objects and signs) is eliminated by preliminary surface model computation. So 
the surface is excluded from the analysis by subtraction of orthophotos created by using the surface model and original 
images. The method can detect a large variety of objects since it is robust to the object shape and is sensitive to the 
abrupt deviation of the object from the surface only. The probability of detection is higher if the obstacle has sharp and 
contrast edges with respect to the surface. The photogrammetric validation shows high possibilities of the detection on 
the base of usual CCD cameras. 
Further development of the method above is connected with improving various techniques for underlying surface 
reconstruction. 
ACKNOWLEDGMENTS 
We would like to thank Prof. Yury Tuflin for his assistance in this project. 
REFERENCES 
F. Raye Norvell, 1996, Using Iterative Orthophoto Refinement to generate and correct Digital Elevation Models 
(DEM's). Digital photogrammetry: an addendum to the manual photogrammetry, ASPRS,1996 
M.Bertozzi and A.Broggi, 1997, Vision-based vehicle guidance, IEEE Computer, vol.30, pp.49-55, July 1997 
S.M.Smith, 1995, ALTRUISM: Interpretation of three-dimensional information for autonomous vehicle control, 
Technical report, http://www.fmrib.ox.ac.uk/~steve 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 1047