'odesy.
squares
Matrix
[ueller,
rithms:
Least
"lando.
uorela,
stration
’RS 29
ions in
Royal
Im.
ISPRS
Im.
PERS
algebra
5-3/3.
m: An
nmetry.
digital
algebra.
im on
er 25.
tomated
ithm of
19-465.
Ly Least
for the
61-72
Three-
Is using
-1389.
(imation
38.
A CONCEPT FOR A NETWORK-BASED DISTRIBUTED IMAGE DATA ARCHIVE
Herwig Rehatschek, Institute for Computer Graphics, University of Technology Graz, Austria
E-mail: herwig@icg.tu-graz.ac.at WWW: http://www.icg.tu-graz.ac.at/-herwig/basicinformation.html
Commission II, Working Group 5
KEY WORDS: Design, Distribution, System, Management, Archiving, Spatial, Database, Networks
ABSTRACT:
Large image archives offering data from planetary missions and the Earth are being created. The searching for data about points and
regions of interest and the request for a specific coverage remain tedious processes. Retrieved full resolution image data stored on a
CD or a tape usually exceed the area of interest by far. The proposed GDSS (Graz Distributed Server System) therefore holds
quicklooks of all available images and surface maps at different levels of detail representing the planet’s surfaces, and supports the
search for full resolution data of just the area of interest. GDSS provides remote access for searching, querying and image processing
facilities via a high-speed ATM backbone. We present a snapshot of the system design and the expected results of the GDSS project:
a functional prototype including software, a network testbed, a series of data experiments and some simulation results. Data man-
agement concepts and storage methods are being illustrated by means of the NASA Magellan data set of planet Venus although we
expect GDSS to also be applicable to Earth based remote sensing data.
1. INTRODUCTION
An increasing number of Earth observation and planetary satel-
lite missions have spawned large image archives. The search for
specific image data by means of additional search criteria such
as mission cycle, time and/or date, geographic location, type of
surface features remains a tedious process due to the current
need of either searching through low quality quicklooks of
poster photo-prints or of dealing with large volumes of source
data that are not organized by the preferred search criteria. That
is why determination of the exact image content is difficult.
Data are often stored in a format that usually exceeds an area of
interest by far.
These experiences are shared by a wide scientific community
and are exemplified in this paper by the image data set of planet
Venus which was produced by NASA's Magellan spacecraft.
NASA's Planetary Data System (PDS) (NASA PDS, 1995)
offers access to the entire image data set from Magellan. We
report on ideas and results in organizing this data set to ease the
access from remote sites, and in the process also solve issues of
data-specific processing software as well as access to specific
remote computing resources. This work results from the
European Magellan. Data Node (EMDN, 1995) which is
organized as part of NASA's PDS under an agreement between
NASA, the Austrian Space Agency and the Institute for
Computer Graphics at the University of Technology, Graz.
During the Magellan mission to Venus about 9594 of the
planet's surface was mapped by the spacecraft's sensors. The
data set consists of SAR images (>5200 orbits) with a total
volume of about 500 GByte. The raw images are strips which
are about 350x220000 pixels in size at a resolution of about 75
meters/pixel. The orbits (polar, north to south) are grouped into
three “cycles defined by the different look angles of the SAR
sensor. Images from cycles 1 and 3 can be used for stereo
processing because of their different look angles, both looking
at the surface from the same side. Cycle 2 was recorded from
the opposite side. To get images of rectangular shape, Full
resolution Mosaic Image Data Records (F-MIDR) have been
created. These are 7000x8000 pixels in size and much more
usable than the F-BIDRs in terms of searching, cropping,
Storing and visually interpreting. The F-MIDRs were only
327
created for about 15% of the surface. The entire surface is
covered by compressed mosaics (C-MIDR) at reduced
resolutions. The desire to offer “Full” resolution, yet ease of
visual interpretation, has led to a project at US-Geological
Survey (USGS) to create a full resolution total coverage of
mosaic images (F-MAPS) in 1995 (Walcher, 1995).
However, this is not supporting the idea of retrieving all images
of a location of interest, and therefore exploiting the
multiplicity of images. We propose that not only searching, but
also processing and retrieving image data like Magellan's, be
based on the Graz Distributed Server System (GDSS). It is
designed for giving a geographically dispersed scientific com-
munity easy and unified access to all planetary data, source
image processing capabilities and computing tools. The basic
GDSS ideas are not restricted to the Magellan data set of
Venus. Data management concepts, access to. parallel
computing resources and network design should be applicable
to Earth based remote sensing data.
2. RELATED WORK
Basic differences exist in handling Earth and planetary data.
Search and retrieval systems on Earth-based image data can
take advantage of many named features, and just about one
third of the surface is covered with land. These advantages do
not exist on a planet like Venus. Therefore a raster image
oriented map has to be used instead of a vector oriented one.
Earth observation with its large spatially organized data has
spawned various systems for image cataloging and remote
retrieval. Examples include GISIS (Graphical Intelligent
Satellite Information System) (Lotz-Iwen, 1995) and VISTA
(Visual Interface for Space and Terrestrial Analysis) (Snyder,
1994), which are designed to give remote access to various
remote sensing data using a special purpose client software.
GISIS supports the user with an intuitive GUI and a detailed
zoomable 2D vector map of the Earth, VISTA additionally
supports a 3D vector representation of Earth. The European
Union's CEO (Center of Earth Observation) was initiated to
develop a European wide support of multi mission data
catalogues. A first spin off is the so called European Wide
Service Exchange (EWSE, 1995). It has some special new
features which allow registered users to search, input, update
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996
