|
|
Conceptual overview of the AVO prototype SWG release
M.G. Allen, Jan 30, 2003
This document is provided as an outline to some of the
concepts employed in the AVO demo. We focus on the
metadata and data model aspects, and try to provide a
snapshot of the ideas that are driving some aspects
of VO-like software development.
Metadata Browsing and Visualization
The AVO demonstration prototype consists of two main
parts: "Metadata discovery and manipulation", and
"Visualization". In the prototype these two concepts
relate respectively to the GOODS data tree in the
server Selection window (metadata browsing), and the
main Aladin style window (visualization).
The metadata browsing capability in this prototype
represents an important first step towards
defining the infrastructure which will enable Virtual
Observatory style operations. It enables browsing of
the metadata of large datasets, in this case the
Great Observatories Origins Survey data, in a way
that allows detailed selection of data at a given
point in the sky.
The visualization component is more familiar to our
everyday use of Astronomical software tools. The AVO
demonstration prototype includes many functions available
in current high level data visualization packages, such
as a multi-plane image stack, colour table manipulation,
RGB colour composition, contours, coordinate grids and
text overlay. The AVO prototype utilizes the CDS Aladin
sky atlas tool as a basis for the visualization component.
New functions in the prototype
The new functionalities on the visualization
side build on the catalog-overlay features for which Aladin
was first designed. In this AVO prototype, multiple catalogs,
may not only be loaded into planes of the image stack,
but they may be manipulated, "filtered", and grouped into
folders (within the stack) in order to operate on the catalog
values. For example, one may select catalog values (of a
photometric catalog) corresponding to various colour criteria,
or any other criteria based on values in the catalog columns.
Another important improvement built into the prototype is
the SED tool, which allows plotting and manipulation of an SED
in a graph window. This represents not only a new function,
but demonstrates that various modular software components
(Aladin from CDS, and the SED tool from ESO) can be integrated
together. Such modularity is expected to play a large role in
VO portals, or visualization components.
Metadata browser, and the Data Model (in the AVO demo and beyond)
What is special and new here is the way the information about the
images (that is the image "metadata"), is "discovered", stored,
and presented. This prototype employs test versions of the
IDHA data model. The data model is essentially a schema, expressed
in the Unified Modeling Language which describes metadata about
images, and their organization. When a request is made on the
database, here the GOODS database, a "view" or "projection" of
that model is obtained, resulting in an XML description of the
database. This XML description is then used to construct the
data tree as seen in the GOODS load window in the prototype.
In this prototype, the query is sent to the the data model (and
the XML description generated) only once per session.
That query is sent when the "GOODS" button is pressed. The
hierarchical tree is built from the XML description, and then
remains static in the browser for the remainder of the session.
Also, in this prototype the user is not offered any options to
change the way the tree is built or displayed.
The concept employed here is however much more general, and can potentially
provide a high degree of flexibility in describing both very large
and very small datasets. In principle the initial request can specify
a radius, such that the XML description that is generated, only
contains data relevant to the region of sky as specified by the
position and radius. That is, the whole database need not be described,
only the relevant portion.
Similarly, the initial request may be on some other set of parameters
(as might be described in the data model), say pass-bands within
given wavelength constraints, observational epochs, or other combinations
of parameters.
Also worth emphasizing here is the fact that the hierarchical tree
description of the database (as used here for GOODS), that was
dynamically built from the data model, is only one of many possibilities
for describing the data. We chose this one for simplicity, and because of
the general familiarity with directory-tree type structures. Different
"projections" of the data model, may be described by different list
or map like structures, which could be supported in the same framework
as developed for this prototype. This kind of framework, because it
naturally includes metadata, can be used to provide high level data
selection cpabilities such as the image field of view projection
included in the AVO demo prototype, and the method of identifying
(and selecting) all the data available at a given location, on the
sky, or within a region of a given parameter space.
How easily may such a framework be applied to other data sets?
In principle, a dataset which is mapped onto the IDHA data model
can be supported in the same framework. This is however a prototype,
and given the rapid development of data models in the international
VO community, the data model as used here is presented as a
demonstration of the concept. Also, there is much work to be done
in order to describe other data such as catalogs and spectra, so
that these may also be incorporated into a framework like the
prototype outlined here.
|
|
Edit menu
