on the support function of the set, the Vorob’eV mean on the
coverage function and the Baddeley–Molchanov on some distance
function. Other recent definitions of mean set can we found in
Jankowski and Stanberry (2010) and Simo et al. (2004).
Among these particularly important mean set definitions, the
Baddeley–Molchanov is the most flexible because it provides different
results depending on the distance function chosen and so
it can be adapted to each specific application. Additionally Baddeley–Molchanov
mean can be applied for general compact sets
because no restrictions are needed. On the contrary, Aumann mean
is only suitable for averaging convex sets i.e. for random compact
and convex sets. Regarding the Vorob’eV mean, although it is applicable
to non-convex sets, it is not suitable for random sets with
zero area.
Regarding our application, Aumann mean cannot be applied
because many of the sets considered are not convex. Both Baddeley–Molchanov
and Vorob’eV definition of mean set can be used
because our observations are sets with positive area. However,
the results of our experiments did not produce good results using
the Vorob’eV average; indeed, the obtained prototype had a
strange visual appearance which, in some cases, did not even have
female shape. Consequently, we decided to use the Baddeley–
Molchanov definition, despite the difficulty to obtain confidence
regions for this (but also for the other) means. Note that the sample
mean of distance functions converges to the population mean using
versions of the central limit theorems for random processes. However,
the Baddeley–Molchanov assume a threshold of the this sample
mean. The closer result can be found in Jankowski and
Stanberry (2010).