Augustana College in Sioux Falls, South Dakota, has taken steps
to establish regional leadership in parallel processing. We requested
an equipped laboratory from the National Science Foundation to develop
a parallel processing laboratory to be used for academic instruction,
together with basic and applied research.
The advent of parallel processing has opened new frontiers in the
ability for computers to solve previously intractable problems.
Until fairly recently, such parallel processing capabilities were
unattainable except through time-sharing arrangements with large
universities or government-sponsored research laboratories. Massively
parallel computers introduced in the early 1990's are nearly all
based on the same, powerful microprocessors that are used in computer
The Parallel Virtual Machine (PVM) project from Oak Ridge National
Laboratories developed a system that enables "a collection
of heterogeneous computers to be used as a coherent and flexible
concurrent computational resource." This software system can
enable the study of parallel algorithms and the simulation of multifarious
parallel architectures on a cluster of computer workstations.
The computer science department has established a parallel processing
laboratory on its own network switch for the study of parallel algorithms
in general, image processing algorithms implemented in parallel,
and parallel solutions to computer graphics problems.
The specific research being performed using the equipment is the
study of parallel implementation of computer vision and image processing
algorithms for the production of real-time, cost-effective image
processing systems. Cluster analysis is very computationally intensive,
and would benefit greatly by parallel analysis and implementation.
This type of unsupervised classification is frequently utilized
on data available from the U.S. Geological Survey EROS Data Center,
Sioux Falls, SD, which has a strong tie with Augustana College through
a NASA contract to provide logistics and related services to the
Data Center. The applications of the unsupervised classification
include land cover classification of large raster images data including,
for example, the entire Great Plains at a spatial resolution of
1 km, 6 layers of data, over at least 36 time periods.