Foreword Highlight Article Columns & Updates Announcements Departments This month's cover is ...

113 Pages·2008·39.75 MB·English

This month’s cover is provided by Dr. Giorgos Mountrakis, Director of the PE&RS
Intelligent Geocomputing Lab at the State University of
New York College of Environmental Science and Forestry
(www.esf.edu). The imagery complements this Special Is-
sue on “Artifi cial Intelligence in Remote Sensing”. Artifi cial
intelligence methods often borrow elements from art and
October 2008 Volume 74, Number 10
science. The background image contains a 2003 natural
color digital orthoimagery of Syracuse, New York with a PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
pixel size of 1.0 ft. GSD. The orthoimagery is processed The offi cial journal for imaging and geospatial information science and technology
using various artistic spatial fi lters. JOURNAL STAFF
The center image is produced using an expert-based
system on an April, 2000 Landsat scene from Las Vegas, Publisher
NV. The system intelligently balances its complexity to James R. Plasker
adjust to the intricacies of the underlying classifi cation. [email protected]
It selectively and progressively moves from simple clas-
Editor
sifi ers to mathematically complex ones such as neural
Russell G. Congalton
networks and decision trees. The image shows the spatial [email protected]
footprint of each algorithmic approach. The integration of
Executive Editor
numerous competing classifi ers offers advanced classifi ca-
Kimberly A. Tilley
tion capabilities; the resulting binary classifi cation of urban
[email protected]
areas is depicted on the top right image. For more infor-
mation contact Giorgos Mountrakis at gmountrakis@esf. Technical Editor
Michael S. Renslow
edu or visit www.aboutgis.com.
[email protected]
Foreword Assistant Director — Publications
Rae Kelley
1199 Special Issue: Artifi cial [email protected]
Intelligence in Remote Sensing Publications Production Assistant
Giorgos Mountrakis and Anthony Matthew Austin
Stefanidis [email protected]
Manuscript Coordinator
Highlight Article
Jeanie Congalton
[email protected]
1178 Next Generation Classifi ers: 11117788
Focusing on Integration Circulation Manager
Frameworks Sokhan Hing, [email protected]
Giorgos Mountrakis Advertising Sales Representative
The Townsend Group, Inc.
Columns & Updates [email protected]
1183 Grids and Datums — Republic of CONTRIBUTING EDITORS
Djibouti
Grids & Datums Column
1186 Headquarters News
Clifford J. Mugnier
1189 In Memoriam — Andrew [email protected]
Piscitello
Book Reviews
1191 Industry News Bradley C. Rundquist
[email protected]
Announcements
Mapping Matters Column
1190 Call for Papers — 22nd Biennial Qassim Abdullah
[email protected]
Workshop on Aerial Photog-
raphy, Videography, and High Web Site
Resolution Digital Imagery for 11118833 Martin Wills
Resource Assessment [email protected]
1194 New Sustaining Member —
Environmental Research
Incorporated
1258 Call for Papers — MultiTemp
2009 — Fifth International 1193 Who’s Who in ASPRS
Workshop on the Analysis of 1194 Classifi eds
Multitemporal Remote Sensing
1195 Sustaining Members
Images
1197 Instructions for Authors
Immediate electronic access to all peer-
Departments 1200 ASPRS Member Champions reviewed articles in this issue is available to
1212 Forthcoming Articles
ASPRS members at www.asprs.org. Just log in
1185 Certifi cation List
1248 Calendar
1186 Region of the Month to the ASPRS web site with your membership
1278 Professional Directory
1187 Refl ection of the Past 1279 Advertiser Index ID and password and download the articles
you need.
1188 New Member List 1280 Membership Application
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING October 2008 1175
OOccttoobbeerr LLaayyoouutt..iinndddd 11117755 99//1177//22000088 88::2299::5544 AAMM
ifi ers: Focusing on Integration Frameworks
by Giorgos Mountrakis
Looking beyond issues related to framework ar-
chitecture, a challenging task remains: establishing
a methodology for optimal selection among several
competing classifi ers, while preserving the desired
characteristics mentioned above. Work being per-
formed in our Intelligent Geocomputing Lab at the
State University of New York College of Environmental
Science and Forestry has begun to show the necessity
and advantages of such a task. We have successfully
established an expert-based system that segments
a binary multispectral classifi cation (e.g. urban vs.
non-urban areas) into context-specifi c sub-problems
(e.g. urban areas of high brightness vs. soil of high
brightness). In a classifi cation of a Landsat scene
(Figure 2) algorithmic complexity adjusts to problem
specifi cs (Figure 3). We have also automated a pro-
cess using eight different neural networks for urban
sprawl modeling. Both works are currently under
review (for updates and paper availability please
visit www.aboutgis.com). The two aforementioned
efforts are small steps towards unifi ed frameworks,
with substantial work still remaining.
Integration Benefi ts Figure 2. Landsat scene from Las Vegas, NV (April, 2000 - natural color).
The underlying objective behind integration is not to
present yet another single-thread classifi er; instead
we strive to establish a framework for collaborative
algorithms. The appropriate merging of multiple
algorithms offers the following advantages:
Support for algorithmic evaluation by non-experts.
Remote sensing products often act as an additional
input layer for numerous environmental studies
(e.g. hydrology, biology, urban planning). It is
often the case that non-experts have high expecta-
tions from remote sensing products without real-
izing potential sensor, acquisition and classifi cation
limitations. Therefore, there is a clear need to incor-
porate advanced accuracy metrics associated with
remote sensing products that express usefulness
and limitations of incorporated methodologies.
Various works already have realized the benefi ts
of spatially-explicit accuracy metrics (Foody et al.,
1992; Canters, 1997; Steele et al., 1998; Carpen-
ter et al., 1999; Pontius, 2000; Alimohammadi
et al., 2004; Liu et al., 2004; Aires et al., 2004).
Integrated frameworks naturally support variable
Figure 3. Spatial footprint of each selected algorithm within the framework.
continued on page 1180
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING October 2008 1179
OOccttoobbeerr LLaayyoouutt..iinndddd 11117799 99//1177//22000088 88::3300::4411 AAMM
continued from page 1179
accuracy metrics, each associated with a specifi c algorithm within new approach for modeling uncertainty in remote sensing
the framework. For example, in Figure 3 there is an accuracy metric change detection process, Proc. 12th Int. Conf. on Geoinfor-
linked to each of the eight algorithms. matics – Geospatial Information Research: Bridging the Pacif-
ic and Atlantic, 7-9 June, 2004, University of Gavle, Sweden,
Error correction capabilities. Assuming a successful integration
pp. 503-508.
framework where each implemented algorithm is independent of Breiman, L., 1996. Bagging predictors, Machine Learning,
another (i.e. independent plug-ins), future algorithmic revisions 24(2):123–140.
should target algorithms with lower accuracy. Ancillary datasets Canters, F., 1997. Evaluating the uncertainty of area estimates
may be acquired in targeted areas (e.g. high-resolution imagery, derived from fuzzy land-cover classifi cation, Photogrammet-
lidar or census data) and as new scientifi c methods arise incremen- ric Engineering and Remote Sensing, 63(4):403-414.
tal algorithmic improvements can be achieved without sacrifi cing Carpenter, G. A., S. Gopal, S. Macomber, S. Martens, C.E.
Woodcock, and J. Franklin, 1999. A neural network method
existing accurate works. For example, as shown in Figure 3, the
for effi cient vegetation mapping, Remote Sensing of Environ-
decision tree (in yellow) is ~82% accurate while the whole scene is
ment, 70:326-338.
~92% accurate, making the decision tree a prime candidate for re-
Coe., Stefan E., M. Alberti, J.A. Hepinstall, and R. Coburn,
vision. We should emphasize that this is due to both the relatively
2005. A Hybrid approach to detecting impervious surface at
low accuracy and the large spatial footprint.
multiple-scale. Proceedings of the ISPRS WG VII/1 ‘Human
Support for scientifi c collaboration. Frameworks may separate a clas- Settlements and Impact Analysis’ 3rd International Sympo-
sium Remote Sensing and Data Fusion Over Urban Areas
sifi cation task in multiple sub-tasks as discussed earlier. There are
(URBAN 2005), 14-16 March 2005, Tempe, AZ.
no restrictions forcing each sub-task to be tackled by the same
Foody, G.M., N.A. Campbell, N.M. Trodd, and T.F. Wood, 1992.
algorithm or scientist. Collaborative environments can be cre-
Derivation and applications of probabilistic measures of class
ated where scientist specialization is at the foreground leading
membership from maximum-likelihood classifi cation, Photo-
to shared instead of competitive efforts. For example, a scientist
grammetric Engineering and Remote Sensing, 58(9):1335-1341.
in one university could establish vegetation extraction algorithms Hansen, L.K., and P. Salamon, 1990. Neural network ensem-
while another focuses on urban build up identifi cation, with their bles, IEEE Transactions on Pattern Analysis and Machine In-
individual efforts later joined together. telligence, 12:993–1001.
Krogh, A., and J. Vedelsby, 1995. Neural network ensembles,
Computational speed. By design, integration frameworks shine in
cross validation, and active learning, within Tesauro, G.,
large-scale applications, because the benefi ts (as mentioned above)
Touretzky, D., Leen, T., Advances in Neural Information Pro-
outperform the initial cost of establishing the framework. In such
cessing Systems, 7, Cambridge, MA, pp. 238, MIT Press.
environments training, error-correction and simulation speeds are
Liu,W.G., S. Gopal, and C.E. Woodcock, 2004. Uncertainty and
important - think of a yearly update of the National Land Cover Da- confi dence in land cover classifi cation using a hybrid classi-
taset. The ability to train and simulate algorithms in a parallel fash- fi er approach, Photogrammetric Engineering & Remote Sens-
ion will utilize the latest hardware developments and in the future ing, 70(8):963-971.
will allow us to analyze much higher data volumes – the majority Perrone, M., 1992. A soft-competitive splitting rule for adaptive
of which is already waiting to be converted into useful products. treestructured neural networks, Proceedings of the Interna-
tional Joint Conference on Neural Networks, Baltimore, MD,
Summary pp. 689–693.
Pontius, R.G., 2000. Quantifi cation error versus location error in
As algorithmic improvements in remote sensing classifi ers reach
comparison of categorical maps, Photogrammetric Engineer-
their limits, the next natural frontier is the integration of multiple ing and Remote Sensing, 66(8):1011-1016.
approaches into a unifi ed framework. In this highlight article, char- Steele, B.M., 2000. Combining multiple classifi ers: An ap-
acteristics for integrated frameworks are discussed, along with a plication using spatial and remotely sensed information for
demonstration of a classifi cation process and associated benefi ts. land cover type mapping, Remote Sensing of Environment,
Considering that image availability is expected to rapidly increase with 74:545–556.
the recent announcement from the USGS to allow free access to the Steele, B.M., J.C. Winne, and R.L. Redmond, 1998. Estima-
tion and mapping of misclassifi cation probabilities for the-
Landsat archive, integrated approaches offer a unique opportunity
matic land cover maps, Remote Sensing of Environment,
for collaborative systems and science within our fi eld.
66(2):192-202.
Wolpert, D.H., 1992. Stacked generalization, Neural Networks,
Acknowledgements
5:241–259.
The author is grateful to the following programs supporting this work:
NASA New Investigator Program, NSF Geography and Regional Sci- Author
ence Program, and Syracuse Center of Excellence Collaborative Activi-
Giorgos Mountrakis
ties for Research and Technology Innovation (CARTI) Program.
Assistant Professor of GIS/Remote Sensing
Director of Intelligent Geocomputing Lab
References
Department of Environmental Resources and Forest Engineering
Aires, F., C. Prigent, and W.B. Rossow, 2004. Neural network State Uni. of New York College of Environmental Science and Forestry
uncertainty assessment using Bayesian statistics: A remote
Syracuse, NY
sensing application. Neural Computation, 16:2415-2458.
[email protected] - www.aboutgis.com
Alimohammadi, A., H.R., Rabiei, and P.Z. Firouzabadi, 2004. A
1180 October 2008 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
OOccttoobbeerr LLaayyoouutt..iinndddd 11118800 99//1177//22000088 88::3300::4466 AAMM
As a prelude to the ASPRS 75th Anniversary, we are running this column to see if you can identify your ASPRS colleagues from their
experiences in the industry. The following material is an abstracted version of an interview conducted by Charles E. Olson, Jr., ASPRS
Emeritus Member and Fellow, as part of the Eastern Great Lakes Region’s Oral History Project. Some questions were used to stimulate the
person interviewed. From what is printed below, can you identify the person interviewed? The person will be identifi ed next month.
Question: “How, where or when did you fi rst become involved in would be an overwhelming number of images, particularly from satel-
remote sensing and GIS?” lites, that would be pouring down to Earth; that would be put in the
hands of environmental scientists and Earth-resource managers, and
Answer: I don’t have to think very long to give a clear answer to this
it would be an incredible experience to be part of that profession and
question because I remember it as if it happened yesterday. It goes
to help it grow. During the second year of my master’s program, he
back to the summer of 1961, when I was in college and enrolled
encouraged me to join ASPRS.
in the Forestry Program. Between the sophomore and junior year,
Let me jump ahead. When I retired, I had gone from a wet-behind-
forestry students were required to take ten weeks of summer fi eld
the-ears graduate student working with black-and-white aerial photos
instruction. About the middle of July, we were led out into a meadow
to Chief of the USGS’s EROS Data Center, which holds the largest
behind the study hall and the subject of study for the day was for-
civilian archive of aerial photography and satellite imagery in the
est photo interpretation in a fi eld setting. Even before the instructor
World—more than 15 million frames of imagery. I was the director
started to speak, it had a profound effect on me, as to what this tool,
of that center and I would have never gotten there if my professor
at the time called aerial reconnaissance, had to offer. In fact, Chuck
hadn’t pointed me in that direction. I certainly thank him for his vision
Olson clarifi ed for me when the term “Remote Sensing” was coined.
and enthusiasm as a mentor.
It didn’t come until a year later, in 1962.
Question: “Who were some of the more infl uential people in keep-
Question: “What prompted you to join ASP or ASPRS?”
ing you in ASPRS?”
Answer: At that time (1966) it was the American Society for Photo-
Answer: I’ve been a member since 1966 and stayed a member of
grammetry. I was working on a plan for the future. Students in those
the ASPRS for all these years because of the Society’s programs and
days thought it was long-term planning if you could get past the next
activities, but mostly because of the people. My membership afforded
Saturday night. But, my plan, since I played basketball, was to enter
me opportunities to make professional contacts. I was able to meet
the military, go to Offi cer’s Candidate School in the U.S. Navy, become
people, who had similar professional interests, beyond the close
an offi cer in the Navy, and then by prearrangement be stationed in
circle that you have in your school or work environment -- scientists,
Hawaii and play four years of basketball for a Navy all-star team. What
engineers, practicing professionals across the nation. My direct in-
would happen beyond four years, after the Navy, I had no clue.
volvement in ASPRS was mainly on the technical side, as a scientist
The professor who taught me to see stereo in the fi eld was a long-
conducting research, writing papers, getting those papers published
term member of the American Society for Photogrammetry. At our
in the peer-reviewed journal, and working with Jim Case, who had
commencement reception he approached me and made a suggestion:
been the long-term technical editor of PE&RS. I worked with Bill
Why don’t you consider graduate school? And, of course, I asked
French, the long-term Executive Director who approached me when
why? He had a vision as to where he thought aerial photography
I was working at USGS, to put training fi lms together for ASPRS. Bill
and remote sensing was headed; what the future would be. One
also suggested that I be one of the Correspondents to ISPRS, it was
vision was, moving from airplanes fl ying a few thousand feet above
Commission VII at the time and was called Photo Interpretation of
the surface, to high-fl ying airplanes, to satellites that orbit the Earth,
Data. I did that for eight years and made many more contacts in the
and not only have fi lm cameras but electronic sensors on board. He
international community. Those that infl uenced me were too many to
basically convinced me that afternoon – while I was still wearing my
name. I’ve only mentioned a few here, but I think it was the colleague
commencement black robe and black hat – that I should chuck the
interaction that really kept me in ASPRS.
Navy approach and go to grad school. He expected someday there
continued on page 1188
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING October 2008 1187
OOccttoobbeerr LLaayyoouutt..iinndddd 11118877 99//1177//22000088 88::3300::4477 AAMM
BE AN ASPRS MEMBER CHAMPION
ASPRS is recruiting new members and YOU benefit from each new member YOU champion. Not only can you contribute to the
growth of ASPRS, but you can earn discounts on dues and merchandise in the ASPRS Store.
Member Champions by Region as of August 31, 2008
Central New York Puget Sound Steven Steinberg REMEMBER! To receive credit for
Thomas E. Henderson Monika Moskal Wubishet Tadesse a new member, the CHAMPION’S
Ricardo Lopez-Torri Carolyn S. Tate
Rocky Mountain name and ASPRS membership
James Mower Phyllis Ullery
Jeff Walton Lloyd P. Queen Keith T. Weber number must be included on the new
Saint-Louis member’s application.
Central US
Jackson Cothren David Kreighbaum Recruited CONTACT INFORMATION
Columbia River Southwest US 5 through 10 For Membership materials, contact us
Christopher Aldridge Stuart E. Marsh New Members at: 301-493-0290, ext. 109/104 or
Daniel L. Civco
Michelle Kinzel Western Great Lakes email: [email protected].
Brian Miyake
Brian Miyake David Hart Brian Murphy Individuals who want to join ASPRS
Robert S. Peckyno Thomas Lillesand Monika Moskal may sign up on-line at https://asprs.
Mike Renslow Member Champions org/application.
Eastern Great Lakes By number of new
James S. Bethel members recruited
Phyllis Ullery RECRUIT
Recruited from
1 new member, earn a 10% DISCOUNT off your ASPRS DUES
Florida 1 to 4 New Members
and $5 in ASPRS BUCK$.
Tarig A. Ali Christopher Aldridge
Brian Murphy Tarig A. Ali 5 new members, earn a 50% DISCOUNT off your ASPRS DUES
and $25 in
Inter-Mountain LloJaymd eHs. SB.l aBcekthbeulrn ASPRS BUCK$.
Lloyd H. Blackburn
Jackson Cothren
Keith T. Weber
Barry Haack 10 or more new members in a calendar year and receive the
Mid-South David Hart Ford Bartlett Award, one year of complimentary membership,
Marguerite Madden Thomas E. Henderson and $50 in ASPRS BUCK$.
Sorin C. Popescu Michelle Kinzel
Wubishet Tadesse David Kreighbaum All newly recruited members count toward the Region’s tally for the
Thomas Lillesand Region of the Month Award given by ASPRS.
North Atlantic
Kenneth W. Potter Ricardo Lopez-Torri Those elibible to be invited to join ASPRS under the Member Champion
Marguerite Madden
Program are:
New England Stuart E. Marsh
Daniel L. Civco James Mower (cid:122) Students and/or professionals who have never been ASPRS mem-
Robert S. Peckyno bers.
Northern California
Sorin C. Popescu
Steven Steinberg (cid:122) Former ASPRS members are eligible for reinstatement if their mem-
Kenneth W. Potter
Carolyn S. Tate Lloyd P. Queen bership has lapsed for at least three years
Potomac Mike Renslow
Elizabeth M. Smith Yogendra P. Singh ASPRS BUCK$ VOUCHERS are worth $5 each toward the pur-
Barry Haack Elizabeth M. Smith chase of publications or merchandise available through the ASPRS
Yogendra P. Singh web site, catalog or at ASPRS conferences.
Mission Statement
ASPRS CONFERENCE
INFORMATION
The mission of the ASPRS is to
advance knowledge and Abstract deadlines
improve understanding of Hotel information
mapping sciences and to Secure on-line registration
promote the responsible
www.asprs.org
applications of photogrammetry,
remote sensing,
geographic information Is your contact information current?
Contact us at [email protected] or log on to
systems (GIS), and
https://eserv.asprs.org to update your information.
supporting technologies.
We value your membership.
1200 October 2008 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
OOccttoobbeerr LLaayyoouutt..iinndddd 11220000 99//1177//22000088 88::3300::4488 AAMM
"From the depths
of the ocean to the
surface of Mars."
- Dr. Allan Carswell, Founder and
Chairman, Optech Incorporated
The founding vision of Optech Incorporated literally spans worlds.  Optech technology is now applied in every
part of Planet Earth and beyond. Our lidar products can be found on every continent surveying from the air,
land and sea. Now Optech is on Mars as an integral part of the NASA Phoenix Mars Mission. Our technology is
at the core of the meteorological lidar delivered to the northern polar region of Mars on May 25, 2008 by the
Phoenix Mars Lander. A mission enjoying spectacular success, this technology is now revealing secrets about
the Red Planet in an effort to understand its atmospheric processes and habitability for life. For the first time in
history, Optech technology is helping scientists study and monitor the atmosphere of Mars, and in doing so,
helping us understand the forces driving climate change here on Earth.
 By studying other worlds we learn more about ours. By working at the leading edge of space, Optech is able to
drive new innovation into its many products.
Many Products - One Family
CCoovveerr..iinndddd 33 99//1177//22000088 88::3344::1166 AAMM
SOCET GXP®SOFTWARE.
EXPERIENCE THE POWER OF EXTREME ANALYSIS.TM
Experience the power of real-time image
analysis,automated geospatial production,
mapping,and 3-D visualization — all in a
single application. SOCET GXP®enables
interoperability and collaboration among
users and decision-makers at all levels,and
its intuitive,customizable interface provides
direct access to shared databases and
extended capabilities. Image and geospatial
analysis—together for the first time in one
powerful tool. This is just one of the ways
BAE Systems delivers real advantage.
www.baesystems.com/gxp
CCoovveerr..iinndddd 44 99//1177//22000088 88::3344::1188 AAMM
S
p
in
e.in Vo
dd 1 lum
e
7
4
, N
o
. 1
0
, p
p
. 1
1
7
3
-1
2
8
0
P
H
O
T
O
G
R
A
M
M
E
T
R
IC
E
N
G
IN
E
E
R
IN
G
&
R
E
M
O
T
E
S
E
N
S
IN
G
O
c
to
b
e
r
2
0
0
8
9
/1
7
/2
0
0
8
8
:4
1
:2
5
A
M
SSppeecciiaall IIssssuuee:: AArrttiiffiicciiaall IInntteelllliiggeennccee iinn RReemmoottee SSeennssiinngg
PPEE&&RRSS
October 2008 Volume 74, Number 10
y
g
o
ol
n
h
c
e
d t
n
a
e
c
n
e
ci
s
n
o
ati
m
or
nf
al i
ati
p
s
o
e
g
d
n
a
g
n
gi
a
m
or i
al f
n
ur
o
al j
ci
offi
e
h
T
GG
NN
SISI
NN
EE
S
E E
TT
OO
MM
EE
R
& &
G G
NN
RIRI
EE
EE
NN
GIGI
NN
E
C C
RIRI
TT
EE
MM
MM
AA
RR
GG
OO
TT
OO SUNY
HH ESF
P
CCoovveerr..iinndddd 11 99//1100//22000088 22::5522::4477 PPMM
Say hello to
IMAGINE Objective.
Ease the pain of manually creating, updating and
replacing geospatial content.
With IMAGINE Objective, you can automate
workflows that analyze remotely sensed imagery,
even with today’s increased imagery resolution
and mapping scales. Convert to vector features
with minimal post processing. Perform both single
feature type extractions and multi-class “wall-to-wall”
classification.
Only IMAGINE Objective combines inferential learning with
expert knowledge – enabling you to get the most value out of
your imagery assets.
Say hello to the Earth to Business Company at
or
Author Manage Connect Deliver Geospatial Business Systems
CCoovveerr..iinndddd 22 99//1100//22000088 22::5522::5599 PPMM