Faber, Petko 
  
  
  
  
  
  
  
Figure 4: Results for the 1st frame of SEQ 3: a) + b) left 
and right stereo image, c) + d) epipolar stereo images, €) 
+ f) extracted points of interest, g) + h) matched points, i) 
+ j) sketched seat approximation, k) snapshot of the gen- 
erated VRML file (only the driver seat is modeled as the 
passenger seat is classified as occupied). 
  
  
  
  
  
  
Figure 5: Results for the 1st frame of SEQ 2: a) + b) left 
and right stereo image, c) + d) epipolar stereo images, €) + 
f) extracted points of interest, g) + h) matched points, i) + 
j) sketched seat approximation, k) + I) sketched human’s 
head approximation, m) snapshot of the generated VRML 
file. 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 237