Geoscience Reference
In-Depth Information
Fig. 1 Height coded
LiDAR-data within a building
ground plan ( left ), delaunay
triangulation ( right )
Fig. 2 Surface normals of
the delaunay triangulation
( left ), test of the orthogonality
of a surface normal in
reference to a building side
( right )
the building. With this method usually each surface normal is assigned to exactly
one side of the building ground plan. Surface normals which cannot be assigned
to a side, for example due to the dispersive behavior of the laser beam, are uncon-
sidered for further calculations. Thus the surface normals are classified, whereby
the number of classes is determined by the number of lines describing the building
ground plan.
The left picture of Fig. 3 shows the surface normals which are split in 8 differ-
ent-colored classes—in which one of the sides is barely visible, due to its short
length. This is often the case in the land register.
For a realistic roof reconstruction a reduction of these 8 classes is required. If
the building ground plan consists of more disruptions, a thereof equivalent number
of classes exist. Each breakpoint in the line of the building creates another class.
Figure 1 shows that the assumed roof structure consists of two saddle roofs pro-
ceeding into each other—therefore of 4 areas. That is why the number of classes
is being reduced by an appropriate selection. As decision criteria for class reduc-
tion the number of surface normals in a class is used respectively the falling
below a certain threshold value. Classes can be deleted if not enough surface nor-
mals can be assigned to them. Additionally, it should be preconditioned that the
Search WWH ::




Custom Search