Geography Reference
In-Depth Information
Discussion
In the article, we showed the deployment of shape metrics on generalized ground
plans of four buildings in order to evaluate geometry generalization quantitatively.
In most cases, the basic assumption that the shape is being simplified during the
generalization was in most cases confirmed. Thus, general trend of increasing/
decreasing shape metrics values (depending on what shape metric is under the
investigation) as the generalization proceeds was also confirmed. Nevertheless, via
calculating shape metrics it is possible to detect discrepancies at particular gener-
alization levels. When looking at shape metrics graph, steep increase/decrease
between two generalization levels could indicate major changes in the shape. In
other words generalization algorithm significantly reshapes the geometry. If this
change follows the general trend, we can claim that the shape was simplified
properly at particular generalization level. On the contrary, if the change of shape
metrics values does not follow the overall trend, the shape was not simplified. Thus,
the generalization algorithm made the shape even more complicated than at previ-
ous generalization level. It is also important to check the graph for peaks. These
local maximums/minimums indicate that generalization either oversimplify, or
does not simplify the shape at particular level. Suggesting that generalization
algorithm performs as it was programmed, we can identify its suitability for
given geometry type as well as generalization level.
Conclusions
Concerning ground plans shapes, here are some findings in connection with
shape metrics. Interesting graph of shape metrics values has the Petronas
Towers. Generalization of its shape, according to the both shape metrics
values and visual interpretation, did not significantly simplify it until the
simplification tolerance values of 200 m, where the geometry turned into a
rectangle. Thus, chosen generalization algorithm did not have desired effect.
On the other hand, the ground plan of Lund University was gradually and
most significantly simplified during the whole generalization process, of
course with some exceptions. Nevertheless, generalization of Lund Univer-
sity ground plan could be considered as the most effective one. Both Houses
of Parliament and St. Maurice ground plans were generalized gradually too
but within a smaller range in terms of shape metrics values. In the first
generalization steps only small spikes were eliminated and then the overall
shape was simplified. Unexpectedly, in several generalization levels their
shape was not simplify but rather vice versa.
According to the case study, we can recommend which shape metrics are
the most useful for quantitative evaluation of generalization. Basic shape
metrics, such as the shape area and length are easy to calculate and are
showing fundamental properties of the shape generalization. These metrics
(continued)
