FRERE EE PSE SAN RS Bee!
TT
IMAGE NAVIGATION FOR EMERGENCY RESPONSE (INFER)
Loey Knapp, Senior Systems Analyst, IBM, U.S.A.
John Turek, Research Staff Member, IBM, U.S.A.
Patricia Andrews, Project Leader, U.S. Forest Service, U.S.A.
Christopher Elvidge, NOAA
Commission II, Working Group 3
KEY WORDS: Developing Countries, Environment, Forestry, Archiving, Navigation, Technology
ABSTRACT
The technology for collection of data through satellite imagery has made rapid strides over the past decade and projections for the size
of image data archives in the next decade are overwhelming. The technology to translate these images into information that is usable,
easy to access, and timely has lagged that of collection, resulting in a data processing and distribution bottleneck. This paper addresses
developments in the provision of image data over the Internet and associated compression, image search, and user interface technol-
ogy, all aspects of the process to translate data into useful information. Application to emergency management in general, and wild-
land fire in specific, is discussed.
INTRODUCTION
Efficient prediction, detection, and response to emergency situa-
tions such as wildland fires requires a substantial database from
which to draw essential information, for example vegetation
coverage and fuel type, adjacent structures and urban areas, and
evacuation or attack crew transportation corridors. Assessment
of fire potential is enhanced by accurate information on seasonal
dryness and change in forest composition. For effective planning
and communication this information must be up to date and
accessible to a wide range of users. To be useful on a global
scale, additional issues arise such as access time, cost, and cross-
country applicability.
Spacecraft imagery has long held the promise of global data
acquisition; by providing up to date global coverage this data
should provide an effective means for augmenting the informa-
tion available through other databases. The large volumes of
data and the physical volume ofthe data which is currently being
collected from a multitude of platforms is already enormous and
projected to grow ten- fold in the next five years. For example,
the instruments on the first two Earth Observing System (EOS)
platforms expected to be launched in 1998 and 2000 will gener-
ate data at a rate of over 281 GB/day with most of this data tak-
ing a form that can be represented as digital imagery. To date the
image analysis process has proved to be a bottleneck with the
result that the amount and quality of information extracted from
this data has not fully met expectations. The challenge is to pro-
vide solutions in the areas of data analysis, storage, retrieval and
dissemination that will help alleviate this problem.
This paper describes the research taking place under the Image
Navigation for Emergency Response (INFER) project to address
these issues. INFER, a subproject under the G7 Global Emer-
gency Management Information Network Initiative (GEMINI) is
a collaborative research effort comprising IBM, the USDA For-
est Service Research, and the Canadian Forest Service Research.
The purpose of the INFER project is to apply new techniques in
signal processing, visualization and data management to con-
struct an environment for interactive navigation and exploration
of large spacecraft image data sets over a network and within the
specific domain of wildland fire mapping and response.
The INFER research is addressing the need to build an infra-
structure that can effectively support search, storage, retrieval,
and transmission of non-traditional data, e.g. images. Unlike
conventional text-based digital libraries and databases, search of
image libraries cannot be realized simply through the search of
text annotations, or metadata. This is because image data is
extremely rich in detail and it would be difficult to provide for
automatic annotation of each image without human intervention.
The INFER team is exploring mechanisms for extracting mean-
ingful information from images through content-based search.
Important components of this work include: 1) the implementa-
tion of a variety of content-based search algorithms, e.g. texture
and shape; 2) application of these algorithms to compressed
data, and 3) data access and dissemination over the Internet.
While this technology may eventually have value in many prob-
lem domains, the INFER project will focus on the development
of the content-based search techniques towards the identification
and mapping of wildland fires.
TECHNOLOGY
Content-based Search
The typical search approach to a large image archive is through
the metadata. The flexibility of the search depends on the extent
of the metadata but common search keys are latitude/longitude,
date, and path/row. Metadata information can be extensive but it
must be created a priori to the search. Thus searches based on
metadata keys work well if the ways in which users want to
explore the data have been anticipated. However, there is little
room for ambiguity and the desire to limit storage of metadata
may radically limit search alternatives.
192
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996
Oui
sea;
the
too!
rith
acte
per
tion
are
For
arcl
new
cate
hav
unli
of ti
Suc
An
com
rithi
shay
sive
agg
As
libre
ann
it is
witk
of w
are 1
will
The
inge
extr:
the i
sear
Extr
extr:
inde
imag
feat
ture
the i
feat
for e
only
ori (
store
not t
ture:
toda
