CC BY
78
ИСПОЛЬЗОВАНИЕ ГЕОПРОСТРАНСТВЕННЫХ ДАННЫХ, ПОЛУЧЕННЫХ ПОСРЕДСТВОМ КРАУДСОРСИНГА, ДЛЯ ПРИМЕНЕНИЯ В СЦЕНАРИЯХ ОТВЕТА НА ЧРЕЗВЫЧАЙНЫЕ СИТУАЦИИ
Евгений Левин
Мичиганский технологический университет, Институт технологии, США, 1400 Townsend drive, Хоутон MI 49931, зав. кафедрой прикладной геодезии, доктор наук, сертифицированный фотограмметрист, тел.+1 (906)487-2446, e-mail: [email protected]
Кришна Мёрфи
Мичиганский технологический университет, Институт технологии, США, 1400 Townsend drive, Хоутон MI 49931, аспирант программы «Интегрированые геопространственные технологии», тел.+1 (906)231-5220,e-mail: [email protected]
Вэйн Толоза
Фирма “Фьюче Кансептс”, США, г. Сан-Димас, Калифорния, президент, тел.+ 1 (909) 5936705, e-mail: wtolo sa@futurec. net
Александр Гречищев
Инновационный научно-образовательный центр “Геомониторинг” Московского государственного университета геодезии и картографии (МИИГАиК), Россия, г. Москва, директор, д.т.н., тел. (916)685-7744, e-mail: [email protected]
В статье описываются текущие ислледования по сбору и анализу данных краудсорсинга для применения в сценариях поанирования и поддержки решений при чрезвычайных ситуациях. Целью исследований является разработка методов, алгоритмов и открытого прораммного продукта для извлечения геопространственной информации с анализом достоверности из текстовых сообщений и изображений размещенных в социальных сетях. Описываемый в статье проект имеет также образовательный компонент.
Ключевые слова: краудсорсинг, чрезвычайные ситуации, геопростанственные
киберинфраструктуры.
СROWDSOURCED GEOSPATIAL DATA DEPLOYMENT IN EMERGENCY SITUATIONS RESPONSE APPLICATION SCENARIOS
Eugene Levin
Michigan Technological University, School of Technology, USA, 1400 Townsend drive, Houghton MI 49931, Chair of Surveying Engineering, Doctor, Certificated Photogrammetrist, tel. +1 (906)487-2446, e-mail: [email protected]
Krishna Murthy
Michigan Technological University, School of Technology, USA, 1400 Townsend drive, Houghton MI 49931, Graduate student in Integrated Geospatial Technology, tel.+1 (906)231-5220, e-mail: [email protected]
Wayne Tolosa
Future Concepts I.S. Inc., USA, 675 West Terrace Drive San Dimas, CA 91773, President and CEO, tel.+1 (909) 593-6705, e-mail: wtolo sa@futurec. net Alexander Grechishchev
Moscow State University of Geodesy and Cartography, Russia, Moscow, Gorochovskiy pereulok, 4, Moscow Geomonitoring Innovative Research and Educational Center, Director, Doctor, tel. (916) 685-7744, e-mail: [email protected]
The paper describes ongoing research effort on crowdsourcing data obtaining and analysis for emergency situation awareness and planning support. Research is challenged on algorithms and open-source software development for extraction with probability analysis of textual and imaging geospatial information from social networks. Described research activity has also educational component.
Key words: crowdsourcing, emergency situations, geospatial cyberinfrustructures.
1. INTRODUCTION
In today’s world, the challenges related to efficient decision support during anthropogenic and natural environmental catastrophes, such as floods, hurricanes, earthquakes, oil spills, terrorist attacks, and others, are associated with obtaining and processing of the vast amount of various geospatial data. This data are deployed in all stages of emergency situations scenarios including: alerts, response, recovery, postdisaster analysis followed by risk mitigation strategy development. One of the promising research fields is feasibility investigation and efficiency estimation of deploying so called crowdsourced geospatial data in application scenarios mentioned above. Crowdsourcingis defined per [F.US,2013]as a process in which individuals gather and analyze information and complete tasks over the Internet, often using mobile devices, such as cellular phones. Individuals with these devices form
interactive, scalable sensor networks that enable professionals and the public to gather, analyze, share, and visualize local knowledge and observations and to
collaborate on the design, assessment, and testing of devices and
results[Heipke,2010]. Related terms include volunteering geographic information, community remote sensing, and collective intelligence.
Nowadays, crowdsourcing plays a major role in creating information-rich maps, collecting geolocalized human activity, and working collaboratively. The convergence of sensing, communication, and computation on single cellular
platforms, and the ubiquity of the Internet and mobile web have allowed maps to be enriched with a variety of data. Early applications included traffic information collected from smartphones, available today from numerous companies (Google, [INRIX], [NAVTEQ], [Waze], [BeatTheTraffic]). The concept was soon extended to enriching maps with other user generated content, either through location-based services or posting from public records. Examples include maps of crime in Oakland,geolocalized real estate data (e.g., [Zillow]), photographic geolocalized postings (e.g., [FLICKR]), pedestrian and sports GPS traces (e.g., Nokia Sports Tracker [NOKIA] ).
The explosion of location-based services has led to the emergence of users sharing personal informationtowards social networks (e.g., Facebook,Twitter), professional information (e.g., LinkedIn), location (e.g., presence in a restaurant, at a landmark location; FourSquare), and social network activities (e.g., placing Facebook
activity on maps; Loopt). This new information complements traditional cell tower information, which is already used in operational contexts, by enriching available feeds using attributes disclosed knowingly or not, willingly or not, by the user.
Finally, new concepts of collaborative work are emerging. Wikipedia created a completely crowdsourced encyclopedia on a voluntary basis. It was followed by numerous services provided by volunteers, such as Facebook translation and [Yahoo!Answers]. Amazon’s Mechanical Turk[MTURK] enables workers to remotely perform tasks at a distributed and large scale for money. This model represents a new trend in which the crowdsourced workers are active and follow directions. This type of activity has been used successfully in geospatial fields for tagging, identification, labeling, parsing, clustering, and recognition.
Openstreet Map an editable map of the world has been used by numerous companies (e.g., [Waze]) as their backbone mapping system. Openstreetmap.org had a remarkable success following the Haiti earthquake of January 2010, when volunteers worldwide created a new map from donated imagery in a few days[Zook et al,2012].
Research described in this paper is conducted in collaboration of the Michigan Tech University with the Moscow State University of Geodesy (MIIGAiK) and the Future Concepts I.S. Inc. in San Dimas, California is challenged on development methods and technology for integration crowdsourced geospatial data in form of location based messages and images into mapping associated applications. Project is foreseen as a culminating by creation of “map of the future is an intelligent 3D image” [Kwan and Lee,2005] as a spatial enabling of information.Paper describes our ongoing collaborative research effort in more details.
2. PROBABLISTIC MATHEMATICAL APPROACH FOR THE ESTIMATION OF THE EXTRACTED FROM CROWDSOURCED DATA INFORMATION QUALITY AND ACCURACY
Any information extracted from crowdsourced data may have inaccuracies associated with the nature of this data. This inaccuracy may follow from the errors in geocoding of relevant text messages in case of their spatial location analysis. Besides human-source producing this information on computer or mobile platform such as smartphone may submit such a message erroneously or under emotional factors influence. In case of imaging information such inaccuracy may follow from the known errors in image features segmentation (which is known as “ill-posed problem”[Marroquin et.al,1987 ]) and matching those features against geospatial features presented in pre-existing database of the area. For geocoded imagery sensor positions are defined by built-in GPS/GNSS sensors and also have well-known inaccuracies [von Watzdorf&Michahelles, 2010]. For the mathematical abstraction let’s assume that we will compare probability of the crowdsoucingly found event or image feature towards simplistic binary model of “truth” and “false”.In this case of binary simplification, the Crowdsourced Information ExtractingSystem (CIES) ——
operates vectors a and b representing a current geospatial object (GO) and data base
vector (DBV), respectively. When b represents a full data stream, and a represents
only a particular view/state of a particular object, a number of a - permutations with N-number of symbols (vector components), and M-number of “ones,” is
( N ^ N!
WN (m) = (m 1
N!(N - m)! ’
where m!=m(m-1)(m-2).(1).
Each a -permutation represents one specific object’s (for example -camera view); typically, we have:
L = 4N .
In this case, the b -database stores all of the object views, which are compared
to a specific GO view, a , obtained during for example video surveillance with smartphone, such as:
1 0 110 1 0 0
-1- 10 0 1 -1- -0- -0-
Agreement
We can see that in any case of agreement (in the brackets), i.e., “1” with “1,” or ‘0” with “0,” the Boolean sum is:
1+1 = 0, 0+0 = 0 .
Whereas the case of disagreement, we have:
1+0 = 1, 0+1 = 1 .
It could be shown for equal probabilities for “1” and “0” (equal to 0.5) that:
£ N 2 - N = 1 ,
m=1 m!(N - m)!
or
2 N = £ N!
T=1 m!(N - m)!
which represents the number of all possible binary permutations. Typically, for example, for a face-front view of a human being, we have:
N = 80 ,
and a typical situation is almost balanced, i.e.,
For an exact balanced case we can calculate the number of permutations as:
WN CK )
N!
(N2)(N2) .
For N=80, we have:
(80)! 53
W80(40) =---------------= 10
80 (40)! (40)!
The False Negative Rate (FNR) can be calculated from the binomial distribution approximated by Gaussian (normal) distribution in the following asymptotic form [Mesgeneu and Murphy, 1976]:
PN (XO
which is illustrated in Figure 1.
1 e
-yo 22
N (y0) HT~
V2^ y0
Figure 1. Illustration of the FNR for Gaussian distribution.
—^
The FNR determines with what probability we can guess the right a distribution, with variance:
S2 = Npq = N
and mean value:
- N
m = qN = — .
2
In CIES context, we assume to use well-known automated target recognition(ATR) so called Figures of Merit (FOMs) that are based on Signal Detection Theory (SDT) [Macmillan,2002]; where conditional probabilities are applies (see, Figure 2); p(SIS), p(SIN), p(NIS), p(NIN), translated as probability of: hit, False-Alarm-Rate (FAR), miss, and correct-rejection, respectively. In this
m =
2
formation, the conditional probability p(AIB) means; if event (B) is present, then event (A) is detected. For example, p(NIS) is probability of signal/target-detection assuming that only noise-is-present. Thus, p(NIS) means also the probability of False-alarm, or False-Alarm-Rate (FAR). We should observe that p(d/I) is similar to p(SIS). Yet p(SIS) is closer to our streaming video scenario, when, on the basis of single, or few frames, we need to evaluate the probability of target recognition (hit).
HIT FAR
p(SIS) p(SIN)
MISS CORRECT REJECTION
p(nIS) p(nIN)
Figure 2.Table of conditional probability according to the SDT model.
Based on this modeling, we here assumeto establishC/ase of Equivalence (COEs), as shown in Figure 3. In more general than in [Thom,1969] case, in addition to probability of hit,miss, FAR, and correct rejection, we here also may computecross-table probabilities (CTP). DuringCIES development, we are research and evaluate the COE-concept for specific classes of equivalence related to emergence situations response application scenarios, for example such as: fire(A), flood(B), human(C), car(D), and others(E).
GO ^^PRESENT GO detecteDx\ A B C D E
FIRE FLOOD HUMAN CAR NOISE (OTHERS)
A p(AIA) p(AIB) p(AIC) p(AID) p(AIE)
HIT CTP CTP CTP FAR
B p(BIA) p(BIB) p(BIC) p(BID) p(BIE)
CTP HIT CTP CTP FAR
C p(CIA) p(cib) p(CIC) p(CID) p(CIE)
CTP CTP HIT CTP FAR
D p(DIA) p(DIB) p(DIC) p(DID) p(DIE)
CTP CTP CTP HIT FAR
E p(EIA) p(EIB) p(EIC) p(EID) p(EIE)
MISS MISS MISS1 MISS CORRECT REJECTION
Figure 3.Table of conditional probabilities, according to the CIES SDT model, generalized into, for example, five classes of equivalence GO like: fire, flood,
humans, cars, and others.
In addition to traditional simplified signal/noise model [Macmillan,2002], where the conditional probabilities mean; hit, miss, FAR, and Correct rejection, we will estimate also Cross-Talk-Probability (CTP).
3. EXPERIMENTS IN 3D INTELLECTUAL MAP PROTOTYPING
WITH CROWDSOURCED EXTRACTED INFORMATION INTEGRATION
Our experimental efforts towards CIES development are associated with development of intellectual agents operating in social media environments. We are working on creation of extended dictionary with words of interest such as "fire", "flood", "earthquake", etc. The dictionary is foreseen as a scalable knowledge base incremented by data mining. Approach to statistical “weighting strategy” for the results of intellectual agents finding is described in previous section. We work on CIES as a Python toolset with the ability to automatically login on facebook, twitter, youtube, flickr in order to search for the dictionary words using, for example, [Mechanize] Python module. Python in an optimal implementation tool because it is deployed as a scripting language for both ESRI ArcGIS and the Open Source equivalent platforms such as [GrassGIS] and [QGIS].
Another important aspect of CIES as a 3D Intelligent Image is efficient 3D visualization integrating results of geospatial data search towards crowdsourcing environments. Experiments in integration CIES functionalitywith SkylineTerraExplorer 3D visualization toolset [Skyline] allows us to load spatial data from files and databases of various standard government and commercial formats. Particularly at the current stage of research we deployed the following capabilities:
- Refreshing of layer content from a spatial database or file to display the most updated information without going through the Load process.
- Reloading layer content from a spatial database around a new location.
- Re-projecting loaded spatial layers from their native projection to the
current TerraExplorer Pro terrain projection.
- Display geographic information for objects, such as length, area and perimeter.
- Loading and displaying tabular information for objects loaded from Shape, ArcSDE and Oracle Spatial.
- Exporting polygon, line and point objects to Shape file.
- Support “Video-on-Terrain” and live web-cams.
- Loading roof-top layers with absolute height
The most important function that we will need for creation of dynamic terrain features generation will be dynamic event’s manager. Initially we foresee these manager interface as development based on TerraExplorer. Particularly API provides us with following capabilities:
- Create extensions to add support for additional file and database formats.
- Real-time movement of objects for command and control applications.
- Add custom-made objects (e.g., danger zone dome).
- Advanced terrain queries (e.g., terrain profile).
- Combine advanced map displays.
- Combine HTML pages with advanced functionality as part of the TerraExplorer window (e.g., Fly to address).
- Create custom run-time applications, embedding the 3D and information windows as ActiveX components
Samples of preliminary results of CIES deployment as an Intelligent 3D map are given on Figure 4 below.
Figure 4.Sample(Meteorite event simulation) of Intelligent 3D image visualization with integration of crowdsourcing data: a) hit map with facebook posts and tweets b) 3D map with live camera integration; c) tweet and facebook videos integration ; d) simultaneous visualization of dynamic platform (car) and video post from this car.
We are also planning to explore 3D visualization capabilities of free and open source 3D visualization platforms such Google Earth and [NASA World Wind].
4. CONCLUSION AND FUTURE RESEARCH
Initial study indicates the actuality and feasibility of the crowdsoucing data integration during all stages of emergency situation response, analysis and preparation. Long term goals of the described research are CIES toolsets integration with Future Concepts I.S. Inc. Emergency Operation Centers (EOC) products and into multiple geoportals such as Geospatial Data Control System [KHusainova et al, 2013].
Research presented has a significant educational component. Michigan Tech and MIIGAiK students involved in research and development the technology for a crowdsourcing system will be exposed to the various knowledge domains as well as develop skills in computer programming , data visualization, database design and management, operating systems, service-oriented architectures, Internet applications, and the ability to work with various types of data feeds.
5. ACKNOLEDGEMENTS
This study was partially supported by The Ministry of Education and Science of Russian Federation under research grant 14.B37.21.1243. Authors want to express their gratitude to “Future Concepts I.S. Inc.” for the financial support of the current research in the frame of student Digital Mapping Enterprise “Morpheus” project. We also would like to bring our special thanks to the Skyline Software Inc. for providing us trial licensees of their products and IEEE for the research license of San Francisco geospatial data set.
REFERENCES CITED
[BeatTheTraffic]http://www.beatthetraffic.com/
[F.US,2013]“Future U.S. Workforce for Geospatial Intelligence”(2013), Academic press book : http://www.nap.edu/catalog.php7record id=18265 [FLICKR] http://www.flickr.com/ https://pypi.python.org/pypi/mechanize/
[GrassGIS ]http://grass.osgeo.org/
Heipke, C. (2010). Crowdsourcing geospatial data. ISPRS Journal of Photogrammetry and Remote Sensing, 65(6), 550-557.
[INRIX] http://www.inrix.com/
Kwan, M. P., & Lee, J. (2005). Emergency response after 9/11: the potential of real-time 3D GIS for quick emergency response in micro-spatial environments. Computers, Environment and Urban Systems, 29(2), 93-113.
R. F.Khusainova, E.Levin, I.S.Kostrukov,V.V.Gavrilova,(2013), “THE SUBSYSTEM OF DATA ACCESS CONTROL TO THE REMOTE SENSING DATA ON THE BASIS OF THE WEB PORTAL USERS’ GEOSPATIAL POSITION”, in Proc. Of Geosiberia 2013[In Press]
[MTURK] https://www.mturk.com/mturk/welcome [NAVTEQ] http://www.navteq.com/
[NOKIA]http://www. sports-tracker.com/ )
Macmillan, N. A. (2002). Signal detection theory. Stevens' handbook of experimental psychology.
Marroquin, J., Mitter, S., &Poggio, T. (1987). Probabilistic solution of ill-posed problems in computational vision. Journal of the American Statistical Association, 82(397), 76-89.
H. Mesgeneu and G.M. Murphy, The Mathematics of Physics and Chemistry, Robert Krlegu Publ., New York (1976).
[Mechanize] https://pypi.python.org/pypi/mechanize/
[NASA World Wind] http://builds.worldwind.arc.nasa.gov/
[QGIS] http://www.qgis.org/
[ Skyline1http://www.skylinesoft.com/skylineglobe/corporate/products/terraexplo rer.aspx
R. Thom(1969), “Topologic models in biology,” Topology, Vol. 8, pp. 313-335.
[Yahoo!Answers] www.answers.yahoo.com
von Watzdorf, S., &Michahelles, F. (2010, November). Accuracy of positioning data on smartphones.In Proceedings of the 3rd International Workshop on Location and the Web (p. 2).ACM.
[WAZE] http://www.waze.com/
[ZILLOW]http://www.zillow.com/7utm source=google&utm medium=cpc&ut m term=zillow.&utm content=73a0d665-484b-0ae9-93a7-00001c7fb153&utm campaign=Branded%20-%20National
Zook, M., Graham, M., Shelton, T., & Gorman, S. (2012). Volunteered geographic information and crowdsourcing disaster relief: a case study of the Haitian earthquake. World Medical & Health Policy, 2(2), 7-33.
BIOGRAPHICAL NOTES
Dr. Eugene Levin, CP, Program Chair of Surveying Engineering and Assistant Professor at School of Technology at Michigan Tech University. Dr. Levin also directing Integrated Geospatial Technology graduate program. He received M.S. degree in astrogeodesy from Siberian State Academy of Geodesy in 1982 and Ph.D in photogrammetry from Moscow State Land Organization University in 1989. He is UP Michigan regional director for of American Society of Photogrammetry and Remote Sensing (ASPRS). Dr. Eugene Levin is an ASPRS Certified Photogrammetrist. Dr. Levin intensively involved in the Michigan Tech Geospatial Initiative development, and is a geospatial technology expert in the fields of photogrammetry, aerial and satellite imagery, remote sensing, GIS, 3D terrain modeling visualization automated feature extraction, and digital cartography. He has 30+ years of experience in US, Israeli and Russian academy and geospatial industry. He held research and managing positions with several Russian, Israeli and US research,academic institutions and high-tech companies, including: Research Institute of Applied Geodesy, Omsk Agricultural Academy, Rosnitc “Zemlya”, Ness Technologies, Physical Optics Corporation, Digital Map Products, American GNC, and Future Concepts I.S. Inc. He has served as a Principal Investigator and Project Manager in multiple award-winning government programs.
Krishna Murthy is a current graduate student at Integrated Geospatial Technology graduate program in Michigan Technological University. He obtained a B.Tech. in Geo-Informatics from Andhra University in 2012. His research is concentrated in integrating 3D Visualization and Geographic Information Systems.
Wayne Tolosa, Founder, President and CEO of Future Concepts I.S. Inc in Los Angeles, California. Mr. Tolosa plays an active role in the management and direction of projects at Future Concepts I.S. Inc. As Project Manager, Mr. Tolosa provides engineering and architectural design of first responder systems for real time situational awareness. For more than 20 years, he taught electronics engineering and computer technology at Citrus College as an adjunct employee. Mr. Tolosa worked in the aerospace arena at the Jet Propulsion Laboratory/NASA prior to founding Future Concepts I.S. Inc. Mr. Tolosa has accumulated 30 years of first responder and incident management experience, holding titles in law enforcement, Search and Rescue and the Fire Service including Deputy Chief of the Baldwin Lake Fire Department, where he was in charge of the Homeland Security Division. He was also the Captain of the Los Angeles County Sheriff’s Department Patrol Reserve Unit and the Captain of the Los Angeles County Sheriff’s Department, San Dimas Search a nd Rescue team. Mr. Tolosa was a featured speaker alongside the Honorable Secretary of Homeland Security, Michael Chertoff and Under-Secretary Jay M. Cohen at the NDIA/Homeland Security Science and Technology Conference (2008) and a Keynote speaker at the National Fire Academy’s discussion on the 21st Century Communications (2010). He has held the position of member of the National Defense Industrial Association’s Board of Trustees in Washington D.C.
Dr. Alexander Grechishchev is Director of Geomonitoring Innovative Research and Educational Center of Moscow State University of Geodesy and Cartography (MIIGAiK). Dr.Grechischev graduated from the Moscow State University of Geodesy and Cartography (MIIGAiK) and qualified as a research engineer with M.Sc.degree in Aerospace Research of Natural Resources research in 1984. In 1991 he got a Ph.D in Remote Sensing from MIIGaiK. After graduating Dr.Grechischev consistently developed his experience holding research and managing positions at multiple organization including the following: research engineer and research scientist at the Special Design Bureau of the Moscow Power Engineering Institute, head of the laboratory at the State Research and Production Center "Nature", the head of the production department and the chief engineer of JSC "Sovinformsputnik", a leading expert in the " DATA + " company, the Deputy Director General of the “Geographic North” company . Since 1997, concurrently with the main work, Dr.Grechishchev used to teach lectures,labs and field trips for the courses "Methods of remote sensing data processing," "Geographic information systems and technology". His scientific and research interests include: remote sensing systems and technologies, methods and technologies for processing of the aerospace data, investigation of Earth's natural resources and the study of the planets of the solar system deploying remote methods; 3D-modeling of objects and terrain, GIS and geoportals development.
CONTACTS
Eugene Levin
Michigan Technological University 4216 EERC Building 1400 Townsend drive Houghton, MI 49931 USA
Tel. +1-906-487-2446 Fax + 1-906-487-2583 Email: [email protected]
Web site: http://www.tech.mtu.edu/Faculty Pages/Eugene Levin.html
Sri Krishna Murthy Hari
Michigan Technological Universirty
221 EERC Building
1400 Townsend drive
Houghton, MI 49931
USA
Tel +1-906-231-5220 Email: [email protected]
Wayne Tolosa Future Concepts I.S. Inc.
675 West Terrace Drive San Dimas, CA USA
Tel.+1 (909) 593-6705 Email: [email protected] Alexander Grechishchev
Moscow State University of Geodesy and Cartography Gorochovskiy pereulok, 4 105064, Moscow Russia
Tel.+7 (916) 685-7744,
E-mail: [email protected]
© E. Levin, K. Murthy, W. Tolosa, A. Grechishchev, 2013
