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


2.6. SPECTRUM: A KBS FOR INTERPRETATION OF IMAGING SPECTROMETER DATA

Borchardt et al.
California Institute of Technology, 1987

As can be judged from the paragraphs above, many of the principal geological expert systems built during the 1980s were focused on the complex and highly qualitiative log analysis task. The system developed by Borchardt et al. (Borchardt 87) addresses another type of problem that has become more and more important to both geology and to the whole domain of natural resource management over the years. Remote sensing by means of aircraft- and satellite-mounted instruments has become an increasingly important means for both exploration as well as for monitoring the earth's surface. If mineralogy and log analysis can be considered analysis of our resources at the microlevel, remote sensing by means of satellites and the subsequent image analysis must be considered a macrolevel approach.

As will be pointed out in later, remote sensing has been addressed for many different purposes. SPECTRUM, built by Borchardt et al., appears to be the first of its kind. The aim of the system was to analyze so called airborne imaging spectrometer (AIS) data. The AIS instrument carried aboard an aircraft would record data over 128 bands in the near-infrared portion of the spectrum for a geographical region of interest. The purpose behind the use of AIS was to manage the remote identification of surface minerals. Various minerals exhibit identifiable characteristics in the spectral range covered by AIS. The recorded image produced by the AIS contains pixels and classes of pixels that can be matched with spectral signatures for tested samples of minerals. However, the richness of the data recorded adds a degree of complexity in the data analysis that is not easily coped with. An AIS data set would typically measure 32 pixels in width, several hundred pixels in length, and 128 levels in depth along the spectral dimension, making a set of more than 2 million individual 8-bit data values. Ordinary pattern recognition techniques are used in order to process the raw image and make it ready for comparison with existing signatures. However, such techniques require substantial expert judgment in order to be applied, due to subtle influences on the recordings caused by atmospheric conditions, calibration of the instruments, and situations where two or more minerals in combination become indistinguishable from a third mineral appearing alone.

The SPECTRUM work aimed at building an intelligent support system for both ground-based as well as on-board operation. Its knowledge included pertinent geology, including descriptions of various minerals, concepts of the relative abundance of minerals, stability under general weather conditions, and interassociations among minerals. It would address materials such as hornblende, albite, anatase, bauxite, epidote, and many more. In addition, it contained a substantial knowledge base about analytic procedures.

Its main representation objects included materials, bands, maps, plots, goals, and plans. In order to achieve its purpose, the developers pursued an architectural approach similar to that used in META/LOG. The system was organized around several knowledge sources called intelligent agents that cooperated in order to reach a common conclusion. The contributing agents were both symbolically as well as numerically based. Another significant aspect of this system was its use of the "mixed-initiative" approach. Both will be explained briefly.

Like many other systems, SPECTRUM included a meta-reasoning unit that could manage different goals according to the development of the problem-solving process. But in addition, it could assign whole plans to achieve these. Goals and plans were applied to direct the ASI data analysis process. Individual goals are appointed and suitable plans that fulfill such a goal are asserted. SPECTRUM carried two plans for each goal. One such plan would be the "combined identification plan." When asserted, it would direct the analysis in the identification of one or more plots, considering both spectral characteristics of plots and the qualitative information describing materials associated with the various plots. Each part of the plan would be associated with one or more agents that would be called upon when that part of the plan was executed.

Like MINDWARE, the group behind SPECTRUM realized the necessity for building a system for practical operations. Among other things, this required the user to stay in control of the analysis if he wanted. The standard rule-based systems up to then had provided little latitude for the user. The "mixed initiative" introduced with SPECTRUM enabled the user not only to override the contributions of the system, but to make it a subordinate partner in terms of control too. As such, SPECTRUM was among the first approaches that combined the standard rule-based approach with a more decision support-like method. One important effect of this was the need for proper truth maintenance. Whenever the user would try to overrule some contribution from a system agent, the chain of dependent assertions would have to be redone to avoid inconsistencies.

The ideas found in SPECTRUM were later adopted independently in other areas such as forestry and ecology. They will be discussed later.


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.