on the
y until
peated
| to a
oducts
tforms
has to
tscape
ftware
s there
Motif
t looks
ascript
yarable
ng to
Iso no
HTML
craft’s
n Data
Graz,
se data
oal an
ise by
for the
still an
proved
© 1996 Wolfgang Walcher
frequently. For a schematic layout see Figure 3. Since Java is
used as programming language, the entire application can be
accessed via the WWW, using the Netscape Navigator without
installing any additional software.
While all standard functions according to 1.3 are necessary for
such a system, the navigation problem is of even more
importance for extraterrestrial data. In this case Venus has
about the same size as Earth. But since there is no water, the
total area of land is three times bigger than on Earth. On the
other hand there are only a few hundred named features, mostly
craters or mountains, which are spread out irregularly on the
Venus surface. This is by far not enough to draw a map that can
be used to localize the content of an image. When searching for
particular topographical features, a detailed representation of
the planetary surface is needed. Up to now this was usually
done using large paper printouts of the digital data.
The ground coverage of planetary image data sets is often much
sparser and therefore it becomes an important criterion when
selecting test areas. Errors and gaps in the data are also
common due to sensor malfunctions or transmission problems.
Visualization of image contents and position combined with
data gaps and sensor characteristics, makes it easy to select test
areas with ‘good’ image coverage or image pairs for stereo
processing (see Fig. 2).
5.2 The dataset
The dataset at the EMDN consists of about 90 Gbytes of image
data and some additional 120 Mbytes of metadata about the
images, spacecraft and sensor characteristics as well as feature
names on Venus. Image data are archived on about 150 CDs,
the metadata are maintained by a relational SQL database. No
spatial indexing scheme is used so far, but an R-tree based data
structure (Guttman, 1984 and Samet, 1990) is under
development to speed up queries for coverage and location.
Details about the dataset are presented in Tables 1 and 2.
Type Number Total Size (Gbytes)
FBIDR 110 15
F/C-MIDR 1250 70
non SAR data 10 1
Table 1: Image data from Venus at the European Magellan
Data Node. For detailed information on the data see
http:/nssdc.gsfc.nasa.gov/planetary/magellan.html.
Type Mbytes Remarks
Image position data 15.0 also of data not at EMDN
Engineering data 110.0 orbit information
Gap Information 1.3 list of all image gaps
Gazetteer 0.1 named surface features
Crater database 0.1 named and unnamed
Table 2: Metadata at the European Magellan Data Node.
201
5.3 Interactive Venus Atlas
This is the main part of the system where all the user interaction
takes place. The main functions are:
Navigation,
Definition of ROI and POI,
Visualization of database content or retrieval results,
Selection of data to preview or order.
The Venus atlas is able to display the entire known surface at 8
different resolution steps from 225 m/pixel to about 20 km/
pixel. The image data are stored in an image pyramid which
was generated from NASA/JPL C1-MIDR data with a pixel size
of 225 m. To ease the user navigation the data have been
resampled from sinusoidal to cylindrical projections to
guarantee that north always points up and south points down.
Because of the distortions of the cylindrical projection the polar
regions had to be treated separately and are stored in an oblique
sinusoidal projection. To reduce the size of the pyramid a radar
speckle removal algorithm . was applied before JPEG
compression was used to store the images. The entire digital
map pyramid is about 600 Mbytes and fits on a single CD-
ROM.
This application takes full advantage of the graphical
capabilities of Java. Only the JPEG compressed image
framelets are fetched from the WWW server and assembled to a
single image map. All user interaction is monitored by the local
Java applets which are also responsible for generating the
vector graphics overlay. As an option program threads running
in the background can preload additional data in advance to
speed up the assembly of the image map.
5.4 Database interface
The interface to the relational database for metadata is also
written in Java and is integrated into the user interface of the
Interactive Venus Atlas. It’s main purpose is to create SQL
queries and to transmit them to the database engine. Retrieval
results are treated in two different ways. The data are first
transformed to a table of human readable text and than all
geometric information is extracted and transformed to a vector
data set for display over the Venus atlas. This is no general
purpose tool and can not be used for other datasets so far.
6 CONCLUSIONS
Traditional query and retrieval programs and interfaces for
access to remote sensing data archives and catalogues have
different disadvantages. Stand-alone programs are usually
dedicated developments for a single archive and are also
available only for a few hardware platforms. HTTP based
systems accessible via a WWW browser do not provide all the
needed functionality due to the restrictions of HTML.
With the recent advent of Java and it’s integration into a WWW
browser like the Netscape Navigator, the above problems can
be avoided. A range of interactive programs is accessible from
any platform via WWW. This concept was used to implement
an Magellan data catalogue and an Interactive Venus Atlas.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996
