2.2.2 TRIANGULAR INTERMESHING AND DENSIFICATION
The major task for PSA is the data capture loop in the defined
area, which consists of the basic triangular intermeshing and the
densification in those regions where the captured points do not
describe the terrain sufficiently to attain the desired height
accuracy.
The basic information with which the terrain analysis and the densi-
fication procedure works is a nearly regular triangular intermeshing
with the specified basic interval as sidelengths. First the bound-
aries are scanned with that distance, and afterwards the triangles
within that borders are then measured.
For the curvature analysis the area is automatically subdivided into
rectangular computing units of fixed width for each densification
step but with a variable length, which depends on the number of
points in that region. The rectangles are always overlapping to
avoid gaps in the model. The most actual information is always taken
into account for the analysis of the next computing unit (Figure 1).
Within the computing units the triangular intermeshing is computed.
Therefore distances between all points in that region are calculated
and sorted depending on their length. Starting with the minimum dis-
tance such sides are eliminated which intersect that side with the
momentaneous minimum sidelength. This automatically generates a
triangle network with a minimum sum of all distances for the edges.
Breaklines and structurelines are always stored at the beginning of
that sorted list of distances so that they form the edges of trian-
gles by default (Figure 2).
For each point the remaining sides form edges of triangles sorting
them depending on their azimuths. If the point belongs to a break-
line the triangular intermeshing is subdivided into two sectors and
two adjusted planes are computed (Figure 3). Depending on the number
of breaklines related to that point the same number of planes is
derived which yields the information on the normals necessary for
the Zienkiewicz interpolation. The analysis of the edges and the
spatial surface of the triangles determines the necessity of densi-
fication and the position of the point to be registered. It also
serves for the prediction of a local extrema inside the triangle,
where the operator is asked for a confirmation emphasises with a
specific acoustic signal and a message. He then has to decide
whether to reject or to accept that prediction. For further analy-
sis an accepted extrema is identified by a horizontal plane and
differs from all other points.
After finishing the densification step for the whole convex quadri-
lateral the current mean square error of height and the maximum
deviation resulting from the difference between predicted and mea-
sured height is displayed on the screen and, if the desired height
accuracy is not satisfied, the next densification step may be
started.
- 13 =
