grad RA or postdoc available in Computational Geometry/Cartography

W. Randolph Franklin wrf at
Thu Dec 9 16:53:51 PST 2004

		available starting Jan 2005

Geologically Correct Terrain Data Structures & Radar Siting

		    W. Randolph Franklin
	      Rensselaer Polytechnic Institute
	  wrf at,

This new DARPA project requires researchers to help us
advance terrain representation and terrain analysis.  New
representations for terrain are necessary for two reasons.
First, the growing volume of data from LIDAR and IFSAR
requires new techniques.  (A 1-meter grid over the Earth
requires $10^{14}$ postings.)  Second, all the existing
techniques have severe shortcomings.  The most important
from the user's viewpoint is that existing techniques make
it easy to represent impossible terrain, but impossible to
represent certain legal and common terrain, such as cliffs.

Our longest-term idea is to avoid weaknesses in current
terrain representation techniques by representing terrain
with a new morphological terraforming operator called
scooping. The resulting terrain is much more realistic than
terrain generated by, e.g., a Fourier expansion.  The second
component of this proposed terrain research addresses the
problem of interpolating terrain from a set of data points
using an overdetermined Laplacian PDE.  The third terrain
representation technique is a Triangulated Irregular Network
(TIN).  TINs have been widely used since Franklin's first
implementation of a TIN in cartography back in 1973.
Because of the large quantity of terrain elevations
produced, e.g., by the SRTM, data compression is necessary.
The compression should be lossy, because the original data
is not perfect and also because of the considerably greater
efficiency of lossy compression relative to lossless.  When
evaluating the quality of a compression, application domain
specific metrics are appropriate.

The first terrain analysis component of this project
concerns efficient include multi-observer siting; see the
poster outside my office (JEC6026).  Drainage net
computation is the next terrain operation.  In principle,
depositing the desired quantity of rain on each point of the
terrain cell, and then allowing it to flow to that point's
lowest neighbor, looks easy.  The reality is different.


Consult my papers and talks on my website for more info on
the topic.

Desirable skills include Math, CS, writing, programming in
C++, linux, GIS, and cartography.

Both RPI students, and others (currently in the US) willing
to transfer to RPI are welcome to apply.

Email your application, in plain text or PDF.  Show how your
background qualifies you.  Start the subject with 'RA


The compgeom mailing lists: see
or send mail to compgeom-request at with the line:
send readme
Now archived at

More information about the Compgeom-announce mailing list