Live edge preserv-
iot parallel to the
system. Because
tions reacts more
tomatic breakline
ection is the der-
> breakline points
1s) in the object
Ihe implemented
information. For
ı method to deal
The investigated areas for the automatic breakline detection
are rectangular areas defined in the object space corresponding
to the marked areas in the digital images above. The details for
the used projects are summarized in the following table.
Open coal- | Mountainous
mine terrain
excavation
image scale 1: 8000 1: 8400
image resolution 15 um 30 um
image type black / white | black / white
DEM size (X/Y) 540m / 640m | 700m / 800m
max./min. spacing 12m / 3n 16m / 4m
min. / max. height in the| 15m/106m |1913m/2199m
object space
number of automatic de- 3911 3170
tected breakline points
Table 4.1: Details of the projects
The first practical applications refer to the current state of the
development (chapter: 3.2). The following figures contain the
final results of the DEM generation process using an adaptive
edge preserving filter. The derived structure information is
superimposed to the DEM's and symbolizes the local surface
areas of maximum curvature in the object space.
Figure 4.7: DEM of open coal-mine excavation area
1
i
!
1
1
1
i
i
I
I
I
i
I
I
i
I
1
i
I
----»| ^
eee n : >
E | £5
1.
Figure 4.8: Breakline area (detail)
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
-—
Figure 4.10: Breakline area (detail)
The representation of the breakline points and their correspon-
ding breakline directions are realized by vectors. The vectors
direct to the local minimum of curvature, hence the direction
of the local maximum of curvature is perpendicular to those
vectors.
The detail figures elucidate the reliability of the breakline
detection. The first geometrical approximation of the break-
lines depends on the spacing of the DEM posts to a high de-
gree. An unsuitable grid width (e.g. grid width too large) can
lead to difficulties due to the vectorization of the point
information.
The algorithm derives a band of vectors in some breakline sec-
tions along the real breakline. Only breaklines parallel to the
object coordinate system provide a clearly breakline descrip-
tion. During to the fact of ambiguities during the vectorization
of the point information we are forced to generalize the gained
result for the next steps of the breakline detection (chapter 3).
But nevertheless, the band of points to the left and right of the
real breakline is a sufficient representation of the breakline for
the surface reconstruction.
To increase the reliability of the breakline detection and to
minimize the model error caused by the quantization of the
DEM we introduced the variable grid width indicated in the
DEM presentations above. The automatic adaptation of the
grid width is controlled using the extracted breakline informa-
tion. In comparison with breakline areas it is not necessary to
