Hi all,
and sincere apologies for cross postings!
I am in the process of starting a phd project on analytical restoration of
paleosurfaces displaced in neotectonics basins in south-central Apennines,
Italy.
One of the key issues at the base of the project is to try and
model in 3- (or 3.5-) D the landforms detected from digital topography,
i.e. to write a system that can discriminate landforms on the DEM,
clusterize them (on the ground of given slope/gradient parameters) and
analytically 'shape-fit' the landform (such as a badland, for instance).
The idea would be to render classes of landforms under an analytical form
which can then be digested by an algorithm to process a DEM, and to cluster
classes of landforms detected on the ground of their shape _and_ of their
analytical description.
I don't exactly know whether/how this would be possibly done, or how has
this been (successfully?) attempted before. Curve fitting may work +/- well
for voice waveforms, and shape fitting may work well for close-range
photogrammetry (a building's deformed wall), but what about naturally
complex, chaotic, fuzzy (?) landforms, that bear a deformational and
erosional history..?
I am thinking about neural networks (for pattern discrimination) and
perhaps finite elements.. to discriminate gradients on the DEM. But I am
very unsure on how to analytically reproduce a landform (provided this is
feasible..) - perhaps with mathematical morphology..? A simple form like an
eroded crater may perhaps be approximated with a spherical integral using a
gauss-like function.. but what about a faulted terrane of which you only
have a few remnants?
Any views, criticisms, suggestions of papers are all most welcome..
Many thanks in advance,
Umberto Fracassi, PhD student
===================================================================
Dip.to di Scienze della Terra Ph.: +39-055-2757522 (office)
Università di Firenze +39-347-6780469 (mobile)
Via G. la Pira, 4 Fax: +39-055-218628
50121 Firenze
Italy Email: [log in to unmask]
http://ewse.ceo.org/anonymous/construct/build.pl/665027
===================================================================
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
