'oduced
oject is
s which
sualiza-
all inte-
ponents
ly one -
rface in
ome
applica-
user
easily
vorking
nderly-
t to con-
black
ality or
search
user in
f the
smart’
oplying
tent of
: much
based
1e fol-
errela-
Mw
MSP.
inter-
e the
features
ng to
tential
and,
rain the
ver the
r three
ary data
plex
sults of
actually
constrain the search by an additional data set such as ‘fires in
grassland areas”.
Unlike SQL, the result of the image query is a spatial display
and, depending on the type of query, the user would also require
information on how the data could be displayed. The user inter-
face must provide the user with the relevant visualization
options. For example, a temporal display could result in a series
of sequential images, a movie format with forward and reverse
buttons, or a single display of multiple small images.
Provision of this complex array of information which changes
depending on the user selections is a user interface challenge.
One option is to base the user interface on a large set of hard-
coded, but regularly updated, tables which prescribe the interre-
lationships across the data and between the data and the display.
This option presents a system’administration problem. The alter-
native is to make the user interface ‘smart’ through a set of data-
base calls, thus providing up to date information for every query.
For instance, if a user selects to query data set A, the system will
search the database for the temporal extent, spatial extent, and
potential search features in data set A and inform the user of the
result. This option is compelling but may have significant per-
formance problems.
APPLICATION OF TECHNOLOGY
INFER Problem Scenario: Wildland Fire Application
IBM and U.S. Forest Service research personnel are jointly
examining the application of these technologies to the area of
wildland fire through the INFER project. The premise of the
research is that effective response to fire is dependent on up to
date, accessible data for prediction, planning and resource allo-
cation. Typically, however, relevant data such as maps of exist-
ing regional fires, fuel type or vegetative greenness, and maps of
populated areas that might be threatened, are not easily obtain-
able nor up to date. The challenge in the area of emergency
response is to provide the technology to extract crucial informa-
tion from images quickly enough to influence the decision-mak-
ing process.
Wildland fire management in the United States is the responsi-
bility of various agencies ranging from federal and state to rural
and private. Fire, however, doesn't recognize administrative
boundaries. This has led to interagency and international cooper-
ation in detection and response to wildfires. In some cases, dis-
patchers and coordinators from various agencies are co-located
to facilitate cooperation and information sharing. Nevertheless
there is room for improvement in tracking fire activity on a
national or regional level across all land ownerships as well as in
archiving historical fire data. National or regional mapping of
fire locations from satellite data through INFER is useful infor-
mation in that it will be up to date and easily accessible across
multiple agencies; that it can be combined with other ancillary
data such as topography, urban areas, and administrative bound-
aries; and that it can be used in conjunction with fire potential
models. It is anticipated that this work might also be extended to
early fire detection applications, especially in countries where a
fire detection infrastructure is not in place.
INFER System
The initial queries supported in the INFER system relate to fire
location at a regional and national level. The data set used in the
prototype is the Defense Mapping Satellite Program (DMSP)
Operational Linescan System (OLS) imagery which provides
daily global coverage with a spatial resolution of eithe r.5 or 2.5
kilometers. The National Oceanographic and Atmospheric
Administration (NOAA) National Geophysical Data Center
(NGDC) has developed forest fire classification algorithms for
the OLS data which rely on detecting areas of active visible and
near infrared emission on the planet surface at night, when solar
illumination is absent. Intensification of the light emissions
results in a data set in which city lights, gas flares, lightning, and
fires can be observed.
The purpose of the INFER project is not to develop new classifi-
cation algorithms but to implement and extend existing algo-
rithms such that they can be used as image search tools. The fact
that NGDC had a process for preprocessing the OLS data and
extracting fire locations made this data set particularly attractive
for the INFER prototype.
The process for locating fire in the INFER system is illustrated
in Figure 1.
IR Band Visible Band
Cloud Detection Visible Light
Extraction
() ' P Q)
Lightning Detection
(3) '
Geolocation
soa
4
(4) | 1 1
Stable Lights
L.] Database
re
Subtraction of
Stable Lights
(5) Y
Aggregation to
Points with Size
and Centroid
(6) v
Fire Data Set
Figure 1: Process for extracting the location of current fires
from the DMSP OLS data.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996
