Brought to you by EarthWeb
IT Library Logo

Click Here!
Click Here!

Search the site:
 
EXPERT SEARCH -----
Programming Languages
Databases
Security
Web Services
Network Services
Middleware
Components
Operating Systems
User Interfaces
Groupware & Collaboration
Content Management
Productivity Applications
Hardware
Fun & Games

EarthWeb Direct EarthWeb Direct Fatbrain Auctions Support Source Answers

EarthWeb sites
Crossnodes
Datamation
Developer.com
DICE
EarthWeb.com
EarthWeb Direct
ERP Hub
Gamelan
GoCertify.com
HTMLGoodies
Intranet Journal
IT Knowledge
IT Library
JavaGoodies
JARS
JavaScripts.com
open source IT
RoadCoders
Y2K Info

Previous Table of Contents Next


6.7. ESS/IIFS

Short et al.
NASA/Goddard Space Flight Center, 1995

Earth System Science (ESS) is a new science promoted by the U.S. Government. In the programs Mission to Planet Earth and Global Change Research, this new science is manifested. ESS describes the earth as a dynamic system where parts interact in response to natural and other processes. The goal of ESS is to understand the earth as a holistic system. Both air, ocean, land, and living organisms constitute the focus. Central to all of this is the use of remote sensing. Like SEIDAM, I2Sadvisor and SPECTRUM presented before ESS will pursue research through object recognition using spectral characteristics. However, the scope is much wider.

NASA has involved itself in ESS. Together with its European counterpart ESA, and Japanese and Canadian interests, the agency has decided to pursue means that will enable integrated studies of the earth's geosphere, biosphere, atmosphere, and cryosphere. Observations needed for these studies will be supported using space-based platforms that will enable recordings of temperatures, ozone depletion, greenhouse effects, land vegetation, and ocean productivity. It will also focus on deserts, forests, and other types of vegetation patterns. Once this is enabled, the data recorded must be made accessible to people with different interests. The task is formidable. Within the 15-year time horizon anticipated, the information system required will produce somewhere around 11,000 terabytes of data. Some of the archive problems will be solved by new mass storage devices. However, the needed accessibility is not guaranteed by this. An Intelligent Information Fusion System (IIFS) is endeavored.

In remote sensing, spectral signatures in the electromagnetic spectrum of known objects are compared to unknown spectral identities. In this way different land forms, vegetation, atmospheric conditions, and artifacts can be recognized in a recorded spectral image. The raw data can cover several bands by means of satellite-mounted sensors. The image processing part involved applies different types of computer-based statistics and classification techniques in order to produce a suitable characterization. The classification effort is positively supported by external data in the form of ground observations, demographics, economics, industrial life, energy consumption, health situation, etc. All this data should be available in a database. GIS is needed in order to produce maps and photo interpretations. The essence of ESS is to produce predictive models about the earth. For this to be possible, the data sampling must cover the entire planet. Another aspect is the need to standardize and integrate recordings taken and produced over a wide time span. Until today, recordings produced by different satellite systems such as Landsat, NIMBUS, and GOES have not followed a common standard. The results from many different platforms and sensors must be coordinated and integrated. In addition to these, more platforms are needed. In order to achieve the holistic goals defined, the ESS initiative must be well managed and pursue the most unified approach possible. The IIFS system will be a critical factor in this. According to Short et al. (Short 95), the purpose of the system is to develop, incorporate, and evaluate state-of-the-art techniques for handling scientific data challenges related to the Earth Observing System required for ESS.

Apart from the satellite systems that produce the raw data, IIFS currently consists of:

  • Mass storage
    • Many devices
    • Semantic modeling
    • Storage hierarchy
  • Object database
    • Object-oriented DBMS
    • Catalog of earth and space meta-data
    • Distributed over many devices
  • Planner/scheduler
    • Expected data rates
    • System and network load
    • Task constraints
  • Meta-data extraction
    • Low-level signal/image processing
    • Neural networks/decision trees
    • Knowledge bases
    • High-performance computing
  • Intelligent user interface
    • Graphical and English querying
    • User customization
    • Planner for analysis assistance
  • IIFS carries two data storing units. One is for the new data received from the satellite units. The other is an object-oriented database customized for fast retrieval of both simple and complex data such as bitmaps.

    A meta-data unit is included in order to increase efficient retrieval of complex data. It contains structured key information about complex data stored. Moreover, it includes spatial data structures called "sphere quadtrees" (SQT). Data originating from particular parts of the earth's surface can be organized according to this structure. One effect is that linearization of the globe surface will be more accurate and thus reflect area projections more precisely. Most of us are familiar with the big, white blob in our school atlas called Greenland. The relative dominance of geographical objects due to 3-D projections onto 2-D maps have always been a source for concern. The SQT structures will minimize these problems.

    The SQT can also be conceived as the global knowledge representation scheme in IIFS. It provides a uniform and recursive structure that is invariant to the size and location of the region represented. SQT can in turn provide a structure called the "hyper cylinder," which organizes all meta-data about satellite observations of the earth or space in the object-oriented database. Hence, a mosaic of results within a region can be efficiently built up. It will be produced by putting together several pieces of data recorded, processed, and stored independently of each other.

    Apart from being able to store complex data structures such as geographic images, the object-oriented database is also more readily maintainable. When data types are known and data are available, the retrieve operation is fairly standard. If the query includes names of unknown concepts, these types can easily be incorporated. When data are not immediately available, the planner module is invoked. This implies that the data requested is not in store, but can be ordered. The planner posts the required result as a goal and tries to define a plan that will achieve this goal. The plan may involve several processing steps, as illustrated by the following:

    • Identify an appropriate scene
    • Identify and allocate machine resources for the given task
    • Isolate the region of interest
    • Compute the algorithms involved
    • Place the resulting image file at an appropriate location
    • Inform the object database of the status and availability of results

    As a human warehouse manager, the object database will then relay the final result to the user. The planner distinguishes between different types of goals. A major distinction exists between hard goals and soft goals. When pursuing a conjunction of goals, as many as possible are attempted. However, if rewards are low, the system reduces its demand according to the status of its goals. Soft goals are first discarded.

    One of the most challenging things addressed in this system is the automated characterization of image content. This is important because images can be indexed according to content rather than less specific keys such as time and geographic area. In IIFS, the substance of each image is defined and summarized. The features contained in an image are held in a characterization vector. Different sensors are assigned distinct recognition objectives. Each one will detect a finite set of objects. Image processing algorithms are specifically created for a sensor and the kind of objects that it will focus toward. All characterization algorithms use clustering techniques. Both supervised and nonsupervised classification can be undertaken. Approaches used include traditional Bayesian classifiers as well as neural networks. Backpropagation variants have been applied. These have produced the highest level of accuracy.

    For more information on this extensive system and the ESS, we refer the reader to the comprehensive article by Short et al (Short 95).


    Previous Table of Contents Next

    footer nav
    Use of this site is subject certain Terms & Conditions.
    Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.