Real-Time Global Data Model for the Digital Earth
Nick Faust, William Ribarsky, T.Y. Jiang, and Tony Wasilewski
Electro-Optics, Environment, and Materials Laboratory
Georgia Institute of Technology
Email: nick.faust@gtri.gatech.edu
Link to full paper
To attack the problem of dealing with increasingly vast stores of global information
for the digital earth, we describe in this paper a general approach to data organization
and real-time exploration based on a novel global hierarchical data model. Our
recent work has revealed that this framework can be quite generally applied to
the earth and anything on it, above it, or even below it. This includes terrain
elevations, phototextures and imagery, maps, buildings, moving or flying vehicles,
weather, and other data. Further, the framework provides a geospatial visual data
mining approach where one can navigate continously from global overviews to high
resolution local views. Because it has an efficient data organization and paging
capability closely coupled to a view-dependent data requesting mechanism (that
is adjustable based on the requirements of the display platform), the framework
is also quite flexible and has been applied to a range of single and networked
systems ranging from a single-processor PC to immersive systems with multiple
projection screens and coupled computers. In this paper we will discuss our geospatial
model, how it fits into an overall framework, and how we have applied it to different
types of data in different environments. We will also discuss our work towards
distributed and Internet-based systems. Our approach starts with a hierarchical
structure for optimal interactivity for data exploration. We use a novel "forest
of quadtrees" with nested coordinate systems for handling global scale terrain
data and permitting timely paging of collections of objects in support of real-time
navigation. We have found that one can effectively employ this hierarchical structure
for a wide range of geospatial data as long as long as methods adapted to the
specific type of data (e.g., terrain versus buildings) are applied at the lower
(detailed) levels. In this paper, we present an analysis of when this should occur
and how large collections of objects, such as buildings and weather clouds, can
efficiently fit into the geospatial hierarchy. The results demonstrate that our
approach is both scalable and general because it is able to handle both large
scale global terrain information and multiple collections of objects (e.g., cities
and weather data) placed around the earth with full interactivity and without
extensive memory load. Further the method shows efficient handling of culling.
Finally, the method shows that levels of detail can be naturally incorporated
to provide improved detail management.