Download Report

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Nonlinear dimensionality reduction wikipedia , lookup

Transcript
GeoInformatics Break-Out Session Report
NSF IDM PI’s meeting
October 10-12, Boston Massachusetts
The field of GeoInformatics is very active and has high relevance to application areas
such as Geographic Information Systems, Location-Based Services and basic sciences
such as astronomy and geology. In our meetings we discussed the recent successes of this
field as well as the challenges that remain. We concluded with a list of action items.
1. Successes – Lessons learned or actions with economic effect
There have been a number of areas where GeoInformatics has had recent impact. For
example, due to the use of the GPS, new location-based services of which we are all
aware have arisen. Automobiles now have online map information. Tourists can rent
systems to tell them about the history and function of buildings they are near. Hikers can
upload map software and use GPS in the wilderness. Online mapping services such as
Mapquest and Terraserver are used extensively.
A GIS industry has arisen. The best known of these is ESRI, which provides a number of
services such as providing maps of property lines for municipalities. An open source
public domain GIS, GRASS is also available.
In the academic sphere, much work on spatial databases and multidimensional indexing
has been published. Currently, commercial DBMSs offer spatial indexing based on this
research.
2. Challenges and Needs
Although there have been successes, there is also much work left to accomplish. For
example, although computer driving directions are now available, they do not take into
account recent events such as bridges out or, even more timely, traffic congestion. In
addition, in following “shortest path” algorithms, drivers are often routed along winding
narrow roads instead of faster highways. This is particularly important for emergency
response directions, disaster recovery and crisis management. If there is an emergency,
we want the most up-to-date traffic and road condition information. The problems of
change detection, modeling and management (such as arise in incorporating traffic
information in driving directions) need better solutions.
The research area of moving objects databases, while vibrant, still has open problems.
Although some proposals for querying moving objects have been put forward, no
standard query language is available. In addition, the problems associated with reactive
behavior for moving objects when changes occur have not been fully solved.
Tools for integration of spatial data which is in different formats (raster, vector, text,
semantics, ontologies, geosensors) need to be developed. Government agencies are very
aware of this problem and are struggling with it. Scientists need to standardize spatial
data to share results. For example, in the recent Microsoft Astronomy project,
(skyserver), Dr. Jim Gray’s group was obliged to convert each astronomer’s data to a
common format by writing a separate program for each conversion.
Integration of spatial data for science is an enormous problem both because the data is
huge and also because it arrives in different formats. It is estimated that less than 5% of
satellite data is used because of these twin problems. Both automated processes for data
analysis (browsing, sampling, real-time generalization, summarization) and applicationspecific data reduction/consolidation techniques of data that are coming from different
and various scientific domains are high priorities.
In addition, much spatial data is textual rather than geometric. This textual data needs to
be extracted from text and correlated with maps and locations. For example, if a
newspaper article says that global warming is causing low-lying islands near Fiji to
disappear, one may wish to correlate this with the geographic location of Fiji and islands
with low elevation in the vicinity. Another example occurs in location based services
which would benefit from integration of timely information about events such as concerts
and exhibitions, which may only be available in textual form.
The general problem of correlating data with location is called georeferencing. One
particular application is in correlating data (say satellite photos) with the USGS
orthophoto quad system. When a feature can be pinned to a specific location on a map,
this is called registration. Problems of registration and georeferencing are largely
unsolved.
Another area needing work is geospatial data mining (pattern and trend discovery and
analysis). Geospatial data mining could be improved with the development of
visualization tools, so human brains (still often better than computer programs) can use
visual representations to discover patterns. Tools for dealing with uncertain and
incomplete data are also needed to improve geospatial data mining.
The recent emphasis on geo-sensor data and stream management is in its infancy. If data
such as traffic density is being gathered in real time from many locations, how is this
location-based data summarized and analyzed?
Specific application areas may need special tools. For example, scientific domains such
as astronomy or biology may need spatial information with unusual requirements.
Chemistry molecule representation, medical informatics such as the required by the new
imaging tools, animation and so forth all have some individual characteristics which may
make them unable to use generic tools.
Last, like all data, spatial and geo data has privacy requirements and ease-of-access
requirements. Some researchers are frustrated in being unable to find meaningful data
sets for experimentation.
As a summary of the challenges above, automation in all things now human-intensive
(data integration, pattern recognition, metadata, easy access, visualization, change
management, summarization, data mining) is the goal. At the same time, each automated
tool must perfectly match the individual requirements of different domains.
3. Actions (not just within NSF but beyond)
Our group suggests the following actions to help achieve the goals of GeoInformatics
Research and to help GeoInformatics Research solve some of our country’s outstanding
social and scientific problems:
o Collaboration with sensors researchers (given the emerging importance of sensor
networks and the huge and growing spatial databases they engender).
o Collaboration on Image/Video segmentation and object tracking (including pattern
recognition), need for automated segmentation of image data.
o Joint initiatives with related sciences (geosciences, climatology, environmental
sciences, transportation etc.).
o More visibility for Geo-Informatics related research in IDM and beyond.