1 Remote sensing image real-time processing for rapid disaster emergency response Dr.Haigang Sui LIESMARS, Wuhan University Oct 23, 2013
2 Outline 1 Major requirements in disaster emergency response 2 Main technical challenges 3 Conclusions 2
3 1 Big spatio-temporal gap between decision maker and actual disaster information 2008 Wenchuan earthquake 2012 Zhaotong earthquake 2013 Yaan earthquake The rapid data acquisition and real-time processing of spatio-temporal data in disaster emergency response is seriously important
4 Sensor network is necessary for comprehensive disaster monitoring InSAR Optical GNSS Hundreds of satellites UAV Airborne Airplane Thousands of aircrafts GBR Meteorological station Seismic Thumper Millions of ground sensors Swimming in Sensors and Drowning in Data David deptula, the U.S. air marshal
5 Rich data, Little information, Poor knowledge For the limitations of rapid data processing, satellite data can be processed in time for less than 10%. How to timely, effectively and rationally integrate dispersed, massive and heterogeneous data is urgent and difficult problem Ingesting all of the data the agency requires remains a major challenge, regardless of how omnipotent an organization is perceived to be. And even once it is collected, analyzing it all in real-time is next to impossible. To watch all the video that currently moves across the Internet in one minute would take five years to watch. And we can t ingest all that data at scale. Meyerriecks the Director of Central Intelligence Agency's Directorate of Science and Technology team
6 2 Main technical challenges Data Wide spatial distribution,and difficult to predict time Irregular and incomplete Scheduled observation in independent space system is difficult to support emergency decision Data are irregular and incomplete in emergency situations. Single-source data processing is much more than that of jointly comprehensive. Information Timeliness requires to be completed in minutes, even in seconds. Multi-sensor Information fusion Title Processing (Near)Real-time Service Online task-oriented service Core problem:real-time continuous spatio-temporal Interpretation using discrete observation data
7 Challenge 1: Irregular and incomplete image data processing in emergency situations Difficulties: Incomplete fundamental data No (little) control points Poor reliability caused by texture-poor regions Irregular, incomplete and non-canonical UAV data processing caused by pose information instable and data errors.
8 Example 1: Irregular data processing Breakthrough automatic image matching of unconventional aviation images in complex terrain and bad weather condition Solve the problem of traditional operation mode s incapability of fast emergency response Simultaneous fly, view and imaging
9 Example 2: Ortho-rectification with few or without control points 影 像 正 射 纠 正 Origin terra-sar Terra-SAR Orthorectified Result Origin SPOT5 SPOT5 Orthorectified Result
10 Example 3: Incomplete data caused by instable small UAV
11 Challenge 2: Real-time processing of integrated space/aerial/ground sensor data Cluster Grid Cloud Multi-Core processor Real-time processing of integrated space/aerial/ground data is just the beginning High-performance computing for Integrated hard/soft and big data Difficulty: How to realize massive spatio-temporal data real-time high-performance computing?
12 Real-time processing is still a long way to go Pixel Factory of France (World-class remote sensing automatic processing system) Pixel Factory of China (DPGrid and PixelGrid) The batch processing techniques for satellite/aerial visible images has been very mature, and the 3D reconstruction of multi-angle camera images is gradually becoming practical. Building Rome in a Day Street Factory s for 3D reconstruction (Paris reconstruction in 2 days, 83%, 30%overlapping)
13 Example : Real-time UAV image data products and change detection Real-time transmit origin image data Change detection in 25 seconds Create DOM in 10 seconds
14 Challenge 3: Multi-sensor data assimilation and information collaborative computing Problem description: Landsat 16 days MODIS 1day Differences in spatial scale Differences in spectral scale SPOT 26 days Aviation mobile observation Continuous observation on the ground X arg max x f [ p spa ( X ), p spe ( Y), p time ( Z), p All ( X, Y, Z)] Differences in time-scale Difficulty: How to achieve assimilation and information fusion with multi-scale multi-temporal/spectral data?
15 Example 1: Oblique remote sensing image processing Image Before Tsunami Registered Tsunami image
16 Building damage extraction based on object-level change detection Object-oriented change detection results
17 Example 2: Optical & SAR image registration SAR images with same area and different side-looking radiometric correction Geometric correction SAR (left) and optical (right) images with the same area Geometric correction with control points Difficulties: Different side-looking and resolution Radiometric and geometric difference Strong speckle noises etc.
18 An automatic optical and SAR image registration method based on VGST & line features 1) Most linear feature based methods use line segments or contours as matching primitive, however, we use line intersection as matching primitive so that conjugate line segment just needs to be collinear and several line segments can provide enough matching primitives. 2) As to matching point features, a new method called VSGT is introduced to improve the poor performance of traditional spectral graph theory (GST) based methods in sensitivity to local optimization.
19 Optical image Line detection of optical image Line intersections of optical image SAR image Line detection of SAR image Line intersections of SAR image
20 Voronoi diagram generated by points from optical image Voronoi diagram generated by points from SAR image Marching result of Voronoi diagram Registration result using MIOI (Chen et al., 2003) Registration result using MIHT (Suri and Reinartz, 2010) Registration result using liner features and Voronoi diagram
21 A simultaneous image segmentation and registration method based on level set Traditionally, segmentation and registration have been solved as two independent problems. However, if we use segmentation result for registration the segmentation methods become very important No matter which segmentation method we use,it is almost impossible to obtain perfect registration result. So it is a good choice to integrate segmentation and registration simultaneously. And this is extremely important to realtime application H. G. Sui, C. Xu, J. Y. Liu, K. M. Sun, C. F. Wen, 2012, A novel multi-scale level set method for SAR image segmentation based on a statistical model, Int. J. Remote Sens., 33(17), pp
23 Example 3: Extraction of quake lake boundary using multi-temporal optical and SAR images FW-2(2 meter)(may,14,2008) Extraction of quake lake boundary of Tangjiang Mountain using SPOT5 (10meters) (May,16,2008) Extraction of lake boundary of Tangjiang Mountain using SPOT5 image before earthquake (Nov,10,2006)
24 Extraction of quake lake boundary of Tangjiang Extraction of quake lake boundary of Tangjiang Mountain Using ASAR (20 meter) (May,22,2008) Mountain Using Radarsat (7 meter) (May,17,2008)
25 Airborne Monitoring Tangjiashan Quake Lake Dam of Tangjiashan Quake Lake ADS40, 0.3m, 28/05/2008 Change of quake lake of Tangjia Mountain with the time
26 Example 4: 3D change detection using optical image and LIDAR data LIDAR data New 3D model generated using UAV images 3D change detection result
27 Challenge 4: Online task-oriented service Disaster emergency response Users needs Geo-spatial Services Service engine Data resources Data finding & binding Information processing & binding Aerospace Information globe Intelligent search Information resources New data acquisition, processing & binding Sensor resources
28 OpenRS: Remotely sensed cloud platform Web applications Web Service Gateway OpenRS-Cloud Job Manager Provide opening architecture for desktop processing, distributed and parallel processing, network service processing. desktop applications Computing nodes
29 Example :Online flood disaster monitoring service Service Reg. Service find Service Chaining Service Running Result Building Search Visualizing Registration Generating geoprocessing abstract 一 of and 个 of geoprocessing 洪 available executing 水 services 淹 model 没 geoprocessing 的 BPEL for 计 result the 算 geospatial 实 service registration 例 GeoGlobe services chains applications center
30 Challenge 5: Onboard real-time processing Onboard processing are mainly concentrated in some data preprocessing algorithms, automatic advanced products generation and information extraction are still very difficult
31 3 Conclusions 4R service ( At the right time and right place to deliver the right information to the right people ) 5R real-time service ( At the right time and right place to real-time deliver the right information to the right people) The world is changing very fast. Big will not beat small anymore. It will be the fast beating the slow. Rupert Murdoch By 2012, 70% of Global organizations will load detailed data into memory as the primary method to optimize BI application performance. - Gartner
BUILDING SPATIOTEMPORAL CLOUD PLATFORM FOR SUPPORTING GIS APPLICATION W.W. Song a, B.X. Jin b,*, S.H. Li b,c, X.Y. Wei b,d, D. Li b, F. Hu e a Department of Geoinformation Science, Kunming University of
Procedia Computer Science Volume 29, 2014, Pages 1246 1255 ICCS 2014. 14th International Conference on Computational Science An Open Framework for Dynamic Big-Data-Driven Application Systems (DBDDAS) Development
20 From Big Data to Meaningful Information with SAS High-Performance Analytics From Big Data to Meaningful Information with SAS High-Performance Analytics Silvia BOLOHAN, Sebastian CIOBANU SAS Analytical
SELECTED TECHNIQUES FOR VEHICLE TRACKING AND ASSESSMENT IN WIDE AREA MOTION IMAGERY A Thesis presented to the Faculty of the Graduate School at the University of Missouri-Columbia In Partial Fulfillment
Int'l Conf. Internet Computing and Big Data ICOMP'15 17 Applying Scalable Machine Learning Techniques on Big Data using a Computational Cluster Dev Dua 1, Sujala D. Shetty 2, Lekha R. Nair 3 Department
MONITORING TRAFFIC BY OPTICAL SENSORS Meffert, B., Blaschek, R, Knauer, U., Reulke, 1 R., Winkler, F., Schischmanow, 1 A. Humboldt-Universität zu Berlin, Unter den Linden 6, 10099 Berlin, Germany 1 German
Chapter 1 Grasping the Fundamentals of Big Data In This Chapter Looking at a history of data management Understanding why big data matters to business Applying big data to business effectiveness Defining
Data Mining with Big Data Xindong Wu 1,2, Xingquan Zhu 3, Gong-Qing Wu 2, Wei Ding 4 1 School of Computer Science and Information Engineering, Hefei University of Technology, China 2 Department of Computer
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING 1 On Understanding Big Data Impacts in Remotely Sensed Image Classification Using Support Vector Machine Methods Gabriele
Mining With Big Data Using HACE 1 R. M. Shete, 2 Snehal N. Kathale 1 CSE Department, DMIETR, Sawangi (M), Wardha, 2 CSE/IT Department, GHRIETW, Nagpur 2 1,2 RTMNU, Nagpur Email: 1 firstname.lastname@example.org,
European Big Data Value cppp - - April 2014 European Big Data Value Strategic Research & Innovation Agenda VERSION 0.7 Contents 1. Executive Summary... 3 2. Introduction... 4 3. Vision for Big Data Value...
The Bloor Group MINIMIZING THE FIVE LATENCIES IN THE ANALYTICS PROCESS How Actian DataFlow Empowers the Data Analyst Robin Bloor, Ph.D. WHITE PAPER Introduction There are several aspects of selecting technology
Data Mining with Big Data e-health Service Using Map Reduce Abinaya.K PG Student, Department Of Computer Science and Engineering, Parisutham Institute of Technology and Science, Thanjavur, Tamilnadu, India
W H I T E P A P E R Cloud Computing Raising Geospatial Technology to the Cloud: Intergraph Strategy for Leveraging Cloud-based Resources Contents 1. Introduction... 1 2. Cloud Computing... 2 2.1. Clustering...
Parallel Computing 29 (2003) 1297 1333 www.elsevier.com/locate/parco Distributed frameworks and parallel algorithms for processing large-scale geographic data Kenneth A. Hawick a, *, P.D. Coddington b,1,
Spatial Cloud Computing -- How geospatial sciences could use and help to shape cloud computing? Chaowei Yang 1, Michael Goodchild 2, Qunying Huang 1, Doug Nebert 3, Robert Raskin 4, Yan Xu 5, Myra Bambacus
Uddin and Al-Turk 1 Airborne LIDAR Digital Terrain Mapping for Transportation Infrastructure Asset Management Waheed Uddin The University of Mississippi Associate Professor of Civil Engineering Director,
Global Headquarters: 5 Speen Street Framingham, MA 01701 USA P.508.872.8200 F.508.935.4015 www.idc.com W H I T E P A P E R B i g D a t a : W h a t I t I s a n d W h y Y o u S h o u l d C a r e Sponsored
BIG DATA: CHALLENGES AND OPPORTUNITIES IN LOGISTICS SYSTEMS Branka Mikavica a*, Aleksandra Kostić-Ljubisavljević a*, Vesna Radonjić Đogatović a a University of Belgrade, Faculty of Transport and Traffic
SuperMap Deskpro.NET Plugin Desktop GIS Based on.net SuperMap Deskpro.NET is a GIS desktop application that integrates 2D with 3D and supports extension development. The key features are as follows: The.NET
A Scheme for the Detection and Tracking of People Tuned for Aerial Image Sequences Florian Schmidt and Stefan Hinz Institute of Photogrammetry and Remote Sensing (IPF) Karlsruhe Institute of Technology
The Case for an Industrial Big Data Platform Laying the Groundwork for the New Industrial Age GE Software gesoftware.com It will take resources and effort, but the Industrial Internet can transform our
DOI 10.1186/s40537-015-0031-2 SURVEY PAPER Open Access Visualizing Big Data with augmented and virtual reality: challenges and research agenda Ekaterina Olshannikova 1*, Aleksandr Ometov 1, Yevgeni Koucheryavy
A Cloud-based System Framework for Storage and Analysis on Big Data of Massive BIMs Hung-Ming Chen a, and Kai-Chuan Chang a a Department of Civil and Construction Engineering, National Taiwan University
JOURNAL OF SOFTWARE, VOL. 9, NO. 9, SEPTEMBER 2014 2427 Network Big Data: A Literature Survey on Stream Data Mining Yi Wu School of Information Science and Technology, Heilongjiang University Harbin, Heilongjiang,150080,
International Journal of Information and Computation Technology. ISSN 0974-2239 Volume 4, Number 1 (2014), pp. 33-40 International Research Publications House http://www. irphouse.com /ijict.htm Big Data