1SPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001
429
Figure 1: Skeleton and circumcentres
Two techniques have been developed, each with its own
physical interpretation. The first, following Thibault and Gold
(2000), uses Blum’s (1967) concept of height as a function of
distance from the curve or polygon boundary, with the highest
elevations forming the crest at the skeleton line.
Figure 2: Elevation model of Fig. 7
Figure 3: Skeleton of a summit
This is illustrated in Figs. 7 and 8, where points on a simple
closed curve are used to generate the crust and skeleton. In Fig.
7, the circumcentres of the skeleton points are given an
elevation equal to the circumradius. The resulting interpolated
model is shown in Fig. 8. This model Is based on the idea that all
slopes are identical, and thus the radius is proportional to the
height of the skeleton point. Of course, in the case of a real
summit as in Fig. 9, the slope would initially be unknown, and
would be estimated from the circumradius of the next contour
level down.
Figure 4: Estimating skeleton heights from circumradii
In the case of a ridge or valley, the circumradius may also be
used, as in Fig. 10, to estimate skeleton heights based on the
hypothesis of equal slopes. The larger circle, at the junction of
the skeleton branches, has a known elevation - half way
between the contours - and may be used to generate the local
slope. The elevation of the center of the smaller circle is thus
based on the ratio of the two radii. For more details see Thibault
and Gold (2000).
Figure 5: Estimating skeleton heights from ridge or valley
lengths
While this method is always available, it is not always the
preferred solution where constant slope down the drainage
valley, rather than constant valley-side slope, is more
appropriate. In a second approach, illustrated in Fig. 11, the line
of the valley is determined by searching along the skeleton, and
heights are assigned based on their relative distance along this
line. This may be complicated where there are several valley
branches - in which case the longest branch is used as the
reference line. This involves careful programming of the search
routines, although the concept is simple. In practice, an
automated procedure has been developed, which uses the
valley length approach where possible, and the side-slope
method when no valley head can be detected, such as at
summits and passes.