NM -—- 
M» 
Vi "QO e= 
N X 
Regine Briigelmann 
  
  
  
reyvalue im: 
(grey image) - second derivatives 
i 
) - homogeneity measure 
I 
breakpointregions 
- hypothesis test 
  
  
      
     
    
  
  
- nonmaxima-suppression 
- thinning 
  
  
  
raster breaklines 
1 
1 
I - raster-to-vector conversion 
I 
raw vector breaklines 
  
  
  
  
  
- splines through x,y-coordinates 
Y - z-coordinate from smoothed surface 
( smooth vector Se | 
Figure 3: Flowdiagram of proposed strategy 
  
  
  
  
  
  
  
   
(a) test statistics 2o (b) breakpoint regions after (c) breakpoints after non- (d) spline interpolated vec- 
hypothesis testing maxima suppression tor breaklines 
Figure 4: Interim results of dike testdata-set 
3.3 Raster-to-vector conversion 
A primitive raster-to-vector conversion first transforms the pixelchain in one-pixel long 3D-vectors which point from one 
pixelcentre to the following pixelcentre. This yields rather zigzag breaklines. A cubic polynomial splining method is 
used for generalizing the lines in the x/y-coordinates. Splines with a vertice distance of three pixels yield rather smooth 
breaklines without a too strong generalization of the lines shapes (fig. 4d). 
4 EXPERIMENTS AND RESULTS 
In this section we describe the used laserdata and the obtained results of the described approach. Beside a qualitative 
description of the results, a quantitative assessment have been performed using photogrammetrically extracted breaklines 
as reference data. 
4.1 Laser data 
The testarea is situated along the river IJssel (The Netherlands). The flight took place in March so that the disturbing 
influence of leaves was minimal on the laser measurements. The dataset was acquired with the helicopter based FLI-MAP 
scanner. The helicopter flew at a speed of 14 m/sec and a mean height of 70 m. With a pulse frequency of 8000 Hz 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 113