Docstoc

Extensions to Geographic Representations UCGIS

Document Sample
Extensions to Geographic Representations UCGIS Powered By Docstoc
					Extensions to representation



         Donna Peuquet
           Penn State
               Research progress
   Data models
    • Conceptual/theoretical
         Spatial representation – agreement…
           •   Objects & surfaces
           •   Discrete / continuous
           •   Things / locations
         Space/time?
           • Recent work seems to be dwelling on the
             philosophical
               What IS an event?

         Multi-representation
           • Agreement on what/when/where
        Research progress

• Implementation
   Much variation on a theme
   Use of existing GIS as a platform

      • ‘new models in old bottles’
                        Needs
   Other tools that allow implementation
    of truly new representations

    =change in paradigm
    Need to investigate the possibilities here
    a) Conceptual
      •   True object representation
          • Not centered on geometry
          • Not restricted to the Relational model
      •   True continuous representation
b.) implementational
   How do we scale up??
     How to implement multi-view with minimal
      redundancy
     Indexing schemes
     Open source database environment
      toolkits (e.g., Berkeley DB)
• Queries?! => Conceptual


  NEED INTERDISCIPLINARY EFFORT
           Recent developments
   Knowledge representation
    • Linked to efforts on ontologies
    • Exploration of implementation tools
         Concept maps, concept graphs, agents,
          neural nets, rule-based
    • Integration with data?
       Auto knowledge acquisition
       Declarative knowledge capture
           The current situation
Visualization &            Database
visual analytics
                           • Static
                           • Tables, vectors or grids
 • Data reside in memory
                              (no semantics)
 • Often don’t consider
   visual and cognitive    • Not intended for
   processes                 driving graphics
                       The need
Visualization &                   Database
visual analytics

 • Data retrieved as               • Space-time dynamics
   needed                          • Integrated with
 • Visual and cognitive              semantics
   processes intrinsic to          • Intended for driving
   the design                        graphics




                 Minimize transformations
             Common, space-time query language
                  Interaction techniques
   REPRESENTATION, broadly viewed
    • [db + kb] + gvis
    • Inexactness
    • Error
    • Multi-scale

    • Cognitive issues
       Cognitively-informed approach
       People/machine task balancing
         Discussion Afterwards
   Abstractions of computer science (CS)
    need to be made practical, workable and in
    context of geographical applications
     • Collaboration with CS
     • Access to proprietary codes? Projects?
   Spatio-temporal query language
   Streamline from rep to viz to knowledge
    discovery
    • From knowledge construction (KC) to
      knowledge use (KU)
   What is an “event”? (in relation to KC and
    KU)
         Discussion cont.
 Rep of PROCESS - all dynamic
 View of world formed by tools? How

  would today’s gamers, cybergeeks
  design a GIS?
 Non-representational approaches?

(e.g., non-spatial data?, need for
  spatialization?)
Spatial Representations and Spatial Data Models
Max Egenhofer


In the area of spatial representations and spatial data models we have seen over the last couple of
years a continuation of a trend that focuses on work in the area of spatial relations, moving objects,
and complex spatial queries. In addition, the modeling of events has become a highly visible
spatio-temporal research topic. Finally, the domain of geosensor networks, and associates spatial
data streams, has made inroads and is poised to become a long-term research issue.

I am looking primarily at papers published in fully refereed key GIScience conferences, because
these papers reflect more accurately recent trends due to the much faster turn-around time from
submission to publication than articles published in the archival journals.

Models of spatial relations continues to be one of the favorite research topics in geographic
information science, as evidenced by the consistent number of papers in GIScience, COSIT, and
SSTD. This year's list of accepted GIScience papers (http://www.giscience.org/submissions.php?4)
includes three-out of 24 total-on projective spatial relations, directed line relations, and trajectory
relations. The 2005 Symposium on Spatial and Temporal Databases (SSTD) featured topological
relations between vague spatial objects, and COSIT 2005 had five papers related to spatial relations
(covering, among others, internal cardinal directions and naming of topological relations). Among
the formalists a new trend can be seen that aims at the study of relation algebras and relation
calculi, with the goal of comparing and integrating the large variety of methods developed for
spatial and temporal relations. Repeatedly the same foundations are used, namely the 9-intersection
and Cohn's RCC-model for topological relations, Freksa's double-cross for orientation relations,
and Goyal's direction-relation matrix for cardinal directions. The cognitive plausibility of such
formal models has been examined primarily by the COSIT community.

The definition of events and the extraction of meaningful events from a myriad of activities is a
relatively new topic that was triggered by the 2002 ACTOR Workshop (Action-oriented
approaches in geographic information science, held in Lucerne, Maine). GIScience 2004 already
included 2 papers on this topic, and more recently further work on event models was presented at
GeoS and COSIT.

A topic related to events is the emerging theme of geosensor networks and spatial data streams.
This is a new domain stimulated by the advent of massive amounts of micro-sensors that are being
deployed in geographic space, either stationary or mobile. The space-time data series that these
sensors pick up form a foundation for near real-time analyses of space-time phenomena. The
computational foundation, in the form of data models, query processing over such data streams, and
the mining for significant changes (i.e., events) is still in its infancy. The 2004 book GeoSensor
Networks (Stefanidis and Nittel) offers a representative collection of recent work.

Since the inception of the European CHOROCHRONOS project in the mid 1990, the topic of
moving objects has been a favorite among spatial and spatio-temporal database researchers. Over
the last few years, the modeling and analysis of trajectories of moving objects has become a regular
theme for papers at GIScience, GeoS, and SSTD.
Particularly the SSTD community is now focusing on algorithms for fast implementations of
queries over massive trajectories.

The 2005 Symposium on Spatial and Temporal Databases highlighted another recent trend relative
to the design of highly specialized algorithms for the fast processing of complex spatial queries.
What started with the k-nearest neighbor problems (e.g., "find the 10 nearest gas station to my
home") has expanded into location-allocation-like queries, such as "find the location to k objects of
type t, such that the average distance to the k objects is minimized" and "find the shortest path
between two locations such that the distance along the path between any two consecutive points
is not greater than a threshold d" and "find the path between two locations such that exactly k
intermediate points are along the path." These types of queries are couched within the domain of
mobile computing, often trying to optimize paths that are constrained by a number of spatial and
non-spatial criteria. The quest is for algorithms that perform fast so that they can work over
massive data sets.

Upcoming meetings, such as GIScience 2006, the Second GeoSensor Network Workshop (Boston,
MA; October 2006), ACM GIS 2006 (http://www.itc.nl/acmgis06/), and SSTD 2007 (to be held in
Boston, MA) are likely to showcase some of the latest advancements in spatial representations and
spatial data models.

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:0
posted:3/16/2013
language:Latin
pages:13