knowledge-centred earth observation final presentation [email protected]
DESCRIPTION
Knowledge-centred Earth Observation Final presentation [email protected]. About ACS. ACS is a private Italian SW engineering company specialised in satellite remote sensing:. Founded in 1979 Headquarters in Rome Research Centre in Matera (since 2001) - PowerPoint PPT PresentationTRANSCRIPT
KEO Final presentation - Esrin 11/12/2008 1
Knowledge-centred Earth Observation
Final presentation
KEO Final presentation - Esrin 11/12/2008 2
ACS is a private Italian SW engineering company specialised in satellite remote sensing:
• Founded in 1979
• Headquarters in Rome
• Research Centre in Matera (since 2001)
• 110 Employees (90% graduates, 11 PhDs)
• Quality System Certification ISO 9001 (since ‘99)
• Certified balance sheet since 1999
About ACS
KEO Final presentation - Esrin 11/12/2008 3
• Integrated Systems
• Software solutions
• Services and Consultancies
In the fields of:
- Earth Observation Satellites Ground Segments
- Environment Monitoring & Management
Applications
- Advanced & Immersive data visualisationTechnology transfer Medical Applications
ACS proposes
KEO Final presentation - Esrin 11/12/2008 4
Ground Segment Technologies
• Payload Ground Segments• Processors (SAR and Optical)• Data Mining • Mission Planning Tools• User Services
ACS Capabilities
KEO Final presentation - Esrin 11/12/2008 5
EO Applications
• Vegetation Monitoring
• Detection of Marine Pollution
• Volcano Eruption Monitoring
• Environmental Risk Monitoring
• Cartography and DEM/DTM
• Desertification Monitoring
• Urban Expansion
• Applied Interferometry
ACS Capabilities
KEO Final presentation - Esrin 11/12/2008 6
ACS Systems in the World
KEO Final presentation - Esrin 11/12/2008 7
• Knowledge-based Information Mining (KIM) tool
– Interactive image collection analysis using Probabilistic Information Mining
– Cooperation of ACS, DLR, ETHZ
– Technology development started in 2001 (ESA TRP) and improved over time
• Knowledge-centred Earth Observation (KEO)
– KIM + rich, graphic, distributed environment for automated Information Extraction
– Cooperation of ACS, DLR, CNES, GTD
– Started in 2005 (ESA TRP)
• Image Information Mining - Time Series (IIM-TS)
– KEO modules for co-registration and Time Series handling
– Cooperation of ACS, CNES, DLR, IGUASSU, JRC, MEEO, SARMAP, UNIPV, UNITN, VTT
– Started in 2007 (ESA TRP)
ACS & IIM
KEO Final presentation - Esrin 11/12/2008 8
The System
KEO Final presentation - Esrin 11/12/2008 9
What is KEO ?
• Distributed Component-based Processing Environment
• Graphically build of Processing Components• Image Information Mining (IIM) Processing with
Interactive Training• Processing results available through OGC
services (WMS, WFS, …)• Processing graphs and/or processing results are
easy to publish in SSE
KEO Final presentation - Esrin 11/12/2008 10
KEO SW architecture
KEO Final presentation - Esrin 11/12/2008 11
KEO HW Architecture
KEO Final presentation - Esrin 11/12/2008 12
FP HW/SW architecture
KEO Final presentation - Esrin 11/12/2008 13
KEO Users
• Domain Expert– Not an expert in remote
sensing– Needs information to make
decision
• Remote Sensing Expert– A scientist with signal & image
processing know-how– Able to build chains of
processing components
KEO Final presentation - Esrin 11/12/2008 14
Live Demos
KEO Final presentation - Esrin 11/12/2008 15
KEO Users Interfaces
• SSE [Domain Expert]• http://keo.esrin.esa.int • KAOS [Domain & Remote Sensing Experts]
– Server Monitor– Image Browser– Image Info Viewer– Product Explorer– World Wind Viewer– FEP Designer
KEO Final presentation - Esrin 11/12/2008 16
Live Demo Glossary
• FEP (Feature Extraction Processor)
• FEP Engine
• FEP Actuator
• Data Flow
• Module [CLI, JAVA, WS]
• Ports
• Connection (Semantic Check)
KEO Final presentation - Esrin 11/12/2008 17
Live Demo NDVI 1
• Problem– Vegetation Detection using NDVI
• Solution– Search in KEO a NDVI modules
• Action– Run the found FEP
KEO Final presentation - Esrin 11/12/2008 18
Live Demo NDVI 2
• Problem– NDVI isn’t good enough
• Solution– Search in KEO for different NDVI modules
• Action– Create a new FEP with the found CLI module
and run it
KEO Final presentation - Esrin 11/12/2008 19
Live Demo NDVI classification
• Problem– Classify the NDVI image to find an easy way
to detect a particular type of vegetation
• Solution– Use an available classification algorithm
• Action– Create a new CLI with the classifier and
integrate it within our FEP
KEO Final presentation - Esrin 11/12/2008 20
Knowledge-based Information Mining
KIM
Live Demo KIM subsystem
KEO Final presentation - Esrin 11/12/2008 21
Live Demo Motivation
• KEO/FEP: a tool for large-scale algorithmic processing– For well-specified phenomena of interest– With well-known solutions
• KEO/KIM: a tool for Exploratory Analysis– For unknown / hard-to-specify phenomena– Via examples [using color, shape, multi-scale
texture for optical, radar sensors]– Interactive KEO processor definition
KEO Final presentation - Esrin 11/12/2008 22
Live Demo Image Browser
• An image archive client:– browse by quicklook
visualization– search by attributes
(e.g. date, footprint...)
– search by content!
KEO Final presentation - Esrin 11/12/2008 23
Live Demo Label Designer
• A rich interaction environment
• Iterative process:– YES/NO examples
via mouse clicks– interface reacts:
red interest areas
• Zoom levels, undo, real-time enhancement...
YES
NO
KEO Final presentation - Esrin 11/12/2008 24
Live Demo Label Designer
• Fast, iterative supervised training across archive
• Can be re-used, refined, extended, managed
• Associated to interactively user-defined textual tags
• Seamless data exchange with archive
KEO Final presentation - Esrin 11/12/2008 25
Live Demo Thematic Maps
• User-defined interest rapid mapping
• Automatic production on-the-fly for the entire archive
• Select + rank relevant images by content
• Can be published on SSE or WCS/WMS
KEO Final presentation - Esrin 11/12/2008 26
Live Demo Product Explorer
• Expert exploratory analysis
• Define new product format combining – Thematic maps– Primitive signals
for further analysis / other users
KEO Final presentation - Esrin 11/12/2008 27
KIM: image search engine
• Application-independent descriptors
• Content simplifier
• Interactivesub-space selection + tagging by labels
Google, Yahoo
• Crawler - data collection / analysis
• Content indexer
• Interactive front end /ranking
“ingestion”
Evolution
KEO Final presentation - Esrin 11/12/2008 28
System Architecture
KEO Final presentation - Esrin 11/12/2008 29
Main componets
• Ingestion subsystem
• FEP subsystem
• KARISMA application
• Reference DataSet
• Ontology subsystem
• OGC services
• KAOS client application
KEO Final presentation - Esrin 11/12/2008 30
Ingestion subsystem
In charge of ingesting products extracted and/or computing:
• information for KIM processing
• features and KIM data
• feature labels and labels maps
• co-registered products
KEO Final presentation - Esrin 11/12/2008 31
FEP subsystem
A data-flow engine in charge of extracting features using Feature Extraction Processors made of:
• Java-coded modules
• CLI programs
• external Web Services
KEO Final presentation - Esrin 11/12/2008 32
FEP subsystem
Made of:• Designer, used to graphically
draw a processing graph• Engine, the executor and
coordinator of the processing graph
• Actuator, the executor of the platform-dependent modules
KEO Final presentation - Esrin 11/12/2008 33
FEP Engine
KEO Final presentation - Esrin 11/12/2008 34
FEP Actuator
KEO Final presentation - Esrin 11/12/2008 35
KARISMA
• The KEO back end application for KIM– Provides access to image archive– Provides access to KIM data– Implements most of the KIM engine– Manages users, projects, …
• J2EE application and some W/S
• Deployed on a glassfish application server
KEO Final presentation - Esrin 11/12/2008 36
KAOS
• KEO client application– KIM client
[KARISMA]– FEP client
[KAFE]– KEO system
management
KEO Final presentation - Esrin 11/12/2008 37
AIR web Service
• Automatic Image Registration– First level of integration of GDAL, but still semi-automatic– Configuration files added for processing different image pairs– Command Line Interface simplification and logfiles centralization
• Deformation Model Management System– XML files based on OGC standards– C++ library for handling a Deformation Model– Test and visualization tools– Some anomalies still to be corrected
• Web Service (AIR-WS)– AIR-WS Java application developed on top of AIR Python scripts– Test and visualization tools
KEO Final presentation - Esrin 11/12/2008 38
Ontology subsystem
Free text search inside KEO by KAOS client
KEO Final presentation - Esrin 11/12/2008 39
Reference Data Set
• Set of reference data used to test and validate FEPs
• Collection of georeferenced Images and any other relevant data (text, pdf, movie, URLs, …)
• Based on GeoNetwork
KEO Final presentation - Esrin 11/12/2008 40
GeoNetwork Features
Users• Search in multiple
catalogues through one website
• Access data from source site
• File download• Data visualization through
web map viewer
Data Providers• Standardized process to
document and publish• Online metadata editor• Different metadata
standards (ISO 19115, ISO 19139, FGDC, Doublin Core)
• Different sharing / security levels
• File upload
KEO Final presentation - Esrin 11/12/2008 41
GeoNetwork Features
GeoNetwork Technology• Java language (platform
independent)• Servlet environment (Tomcat,
Jetty, ...)• XML data representation
(JDOM)• XSL templates for
presentation (XALAN)• JDBC for database access
(runs on any compliant DBMS)
• XML Schema for metadata editing
Catalogue interoperability• Full OGC Catalog v2.0 support• ISO profile
(Reference OGC implementatio)
• Open Archive Initiative support (OAI-MPH)
• RSS and GeoRSS newsfeeds• OpenSearch with GEO
extension• Metadata harvesting &
synchronization
KEO Final presentation - Esrin 11/12/2008 42
OGC services
• WMS and WFS for data extracted by PIM Batch processors
• Based on UMN MapServer
• WMS and WFS exposed to SSE
KEO Final presentation - Esrin 11/12/2008 43
System Evolutions
KEO Final presentation - Esrin 11/12/2008 44
• Technology and Research continues with IIMTS project aimed at extending KEO and KIM capabilities to Image Time Series with the purpose of:
– Identifying and locating changes– Identifying and locating patterns– Tracking and following movement of recurring features
which could represent potentially usefulinformation for specific applications:
– Urban areas monitoring– Agricultural monitoring– Emergency and Security
Further Evolutions in IIM
KEO Final presentation - Esrin 11/12/2008 45
• KEO provides a distributed environment where Feature Extraction Algorithms (including Time Series) can be plugged and used as simple CLI modules (in any programming language).
• Any KEO registered user can deploy his own modules and make these modules usable by others thus optimising development efforts
• By design CLI modules run on processing nodes that can even be physically located on remote sites thus providing an effective solution for a distributed exploitation and processing environment
……………….. And in future ?
Conclusions and Future
KEO Final presentation - Esrin 11/12/2008 46
• Classification of Optical Images (including Very High resolution)
• Tools for Oil Spill and Ship Detection from SAR images
• Binary and Multiclass Change Detection tools from Optical and SAR Images
• Shape Change Detection Tools to point out geometric differences between regions from previous classifications
• Trend Analysis in Temporal Time Series also as an extension of the KIM concept.
From May 2009 KEO will include Processing Components (developed within the Image Information Mining – Time Series project) in support to:
Conclusions and Future
KEO Final presentation - Esrin 11/12/2008 47
• A KEO system will be installed in Romania in the frame of the ESA-ROSA (Romanian Space Agency) agreement.
• ACS continues to study and prototype a second generation of KIM Systems
• A discussion with ESA is undergoing to define a working plan for KIM/KEO enhancements
• An assessment of what should come first in terms of evolutions is needed, so feedback from users would be highly appreciated ………..
………….. What will happen in ‘09 ?
Conclusions and Future
KEO Final presentation - Esrin 11/12/2008 48
• People interested in becoming registered KEO users are invited to contact ESA!http://keo.esrin.esa.int
• For any question please contact:
– ESA for KEO: • Michele Iapaolo [email protected]
– ESA for RDSs:• Andrea Della Vecchia [email protected]
– ACS: • KEO Team [email protected]
Thank you for your Attention
Conclusions and Future
KEO Final presentation - Esrin 11/12/2008 49
Questions?