Ulrich Thoennessen 
  
Figure 3: Scene Frankfurt Airport: a) Map, b) Intensity Data, c) Height Data, d) Depth Map 
Several tie-points with height information can be extracted from the map. Only tie-points on the ground were chosen. 
Correspondences of these tie-points were found in the intensity data. At the related positions in the height data, the 
height values were accumulated in a histogram representation. The histograms showed gaussian distribution with a 
significant offset of the mean value of about plus 8 m in contrast to the map information. Hence, all calculated heights 
in the depth map were corrected by subtracting 8m. After the offset correction, the depth values of the large gassy-areas 
in between the airfield traffic net met about + 2m the true heights. 
The printed map was digitised. A vector map was manually extracted from the scanned map. It contains a layer of the 
generalised representation of the building footprints, restricted to the biggest buildings. The InSAR data is slightly 
rotated compared to the map. For the correction of the rotation and to avoid artefacts caused by foreshortening and 
layover, the vector map is aligned manually on the intensity image. From the aligned building layer of the vector map 
binary mask is generated. 
Figure 4 illustrates the process for a subset of the data in the lower left of the scene around the question-mark-shaped 
large building. To ensure that the buildings are covered entirely by the building mask, the mask is dilated by two pixel 
in each direction. The binary mask is used to separate the depth map in two complementing parts, a “building 
expectation area” and a “non-building expectation area”. 
  
332 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B1. Amsterdam 2000.