Information Mining and Services at ESA

1
Information Mining and Servicesat ESA
S. DElia, M. Iapaolo, A. Della Vecchia EOP-G
Service Support and Ground Segment Technology ESA
- ESRIN, via G. Galilei, Frascati, Italy
2
Table of Contents

• Service Support
• Image Information Mining
• Reference Data Sets
• Ontology Terminology

3
Service Support Projects
European Service Support Environment Enhancements
ESE
Semantic-web for Mediated Access Across Domains
SMAAD
Grid in support to distributed services
SSE-Grid
Enhanced Service Integration Technology
ESIT
SSE Servers and Services
SAS
New Services
NSI
Interfaces
ISAC
MASS Services
MASS-SER-D
MASS Services
MASS-SER-O
Multi Application Support Service System
Environment
MASS-ENV
Multi Application Support Service System
MASS
GSTP
TRP
EOP
4
SSE Concept
SSE Portal

• - Service A
• Service B

• - Provider X
• Provider Y

• Workflow engine activates services
• Data flow outside SSE Portal
• ToolBox (translates terms supports widely used
  legacy stds)

Functions
Directories

• Subscription
• Service Classification
• - Tuneable User Interface
• - Service Characterisation
• Area of Interest
• Monitoring / Control

Workflows
Service A
A1
A2
A3
5
Service Support Environment http//services.eopor
tal.org

• Supports
• Cheap service creation, test, provision
• Services, applications, data remain by SP
• Wrapped as Web Services for publishing
• Easy service chaining (exploit synergies)
• Simple service discovery, fruition
• Services from any domains (e.g. GIS)
• Manual and automatic services (any mix)
• Seamless service evolution
• Characteristics
• Operational since 2006
• Permits new GS interfaces for catalogue, order,
  (ESA EO-DAIL within GSCDA)
• Based on SOA and standards (WS, B2B, B2C, SOAP)
  to solve interface issues
• Test-bed for continuous technological evolution
  HMA, In-Situ, Grid, Semantics
• Multiple installations ESRIN, Senegal, Romania
  (different languages)

6
Example Parallel Catalogue Search
7
Example Information Based Services

• MERIS Cloud-free Subscription Service
  (http//services.eoportal.org)
• Automatically informs about availability of new
  MERIS L1 RR products, with cloud coverage lt user
  defined threshold within user area of interest
• Permits product downloading by clicking on
  provided product identifier

8
SSE Example Services

• MERIS RR Cloud free Product for Cat-1 users
• MERIS Cloud-free Subscription Service
• Reconditioned Landsat Cat-1 Images
• Reconditioned SPOT Cat-1 Images
• Smoke identification by Single Image
• Spectral pre-classification of user image
• Spectrally pre-classified Cat-1 image
• Spectrally pre-classified SPOT Cat-1 image
• SPOT image reconditioning
• SPOT Spectral pre-classification chain
• Urban area seed pixel map from Cat-1 SPOT image
• Urban area seed pixel map from user image
• Water index from Cat-1 image
• Water index from SPOT Cat-1 image
• Water index from user image

• Calibrated and pre-classified Landsat Cat-1 Image
• Calibrated and pre-classified SPOT Cat-1 Image
• Canopy chlorophyll content from Cat-1 image
• Canopy chlorophyll content from SPOT Cat-1 image
• Canopy chlorophyll content from user image
• Canopy water content from Cat-1 image
• Canopy water content from SPOT Cat-1 image
• Canopy water content from user image
• Fire Detection by Single Image
• Greeness index from Cat-1 image
• Greeness index from SPOT Cat-1 image
• Greeness index from user image
• Landsat Image Reconditioning
• Landsat Spectral pre-classification chain

9
Table of Contents

• Service Support
• Image Information Mining
• Reference Data Sets
• Ontology Terminology

10
Image Information Mining Projects
ASIM
Automatic, Semantic Image Information Mining from
Time Series of VHR images
PreQu
AATSR - preclassifier qualification and
multidimensional GUI support
ACIP
ALOS cloud cover and information product
SPA
Support by Pre-classification to Specific
Applications
CARD
Classification Application services and Reference
Datasets
EOLib
The EO Image Librarian EO image and
geoinformation intelligence search engine
KLAUS
KEO demonstrator with models for Land Use
management
SRoKEO
Support to Romanian KEO project
KEI
KIM Extensions and Installations
PIMS-DLR
Partner Information Mining System - DLR
MIMS
MERIS Information Mining System
IIM-TS
Image Information Mining - Time Series
KEO
Knowledge-centred Earth Observation
KIMV
KIM Validation
KES
Knowledge Enabled Services
KIM
Knowledge based Information Mining
11
Knowledge-based Information Mining (KIM)
http//kaos.esrin.esa.int/kaos

• Prototype for interactive image / collection
  analysis(for large / well characterised areas
  Features)

12
KIM Use
13
KIM Output Combined Features Map
14
KIM Primitive Features
15
KEO FEP Graphic Designer
16
KEO Prototype Distributed Architecture
17
KEO Processing Components

• Single image processing
• Calibration (SAR, ASAR, AATSR, Ikonos, Quickbird,
  AVNIR-2, Spot 4/5, Landsat 5/7, MODIS)
• Classification (AATSR, Ikonos, Quickbird,
  AVNIR-2, Spot 4/5, Landsat 5/7, MODIS, SRTM SAR
  DEM)
• Classification for Land Cover (SAR Optical),
  for Urban Class (Landsat), via NN algorithm
• Extraction of MCI / FLH, NDVI / ARVI, Road map
  extraction of Vegetation, Urban or Water maps
  (Quickbird)
• Detection of Oil spill, Ship, Algal bloom, Clouds
  (AVNIR-2) detection of Built-up regions and
  Vegetation map (TerraSAR)
• Interferometric processor, SRTM DEM data
  filtering, SRTM SAR image de-noising, Atmospheric
  error prediction model
• Multiple images processing (from IIM-TS)
• Inter-equalisation / co-registration of time
  series of images Pan-sharpening Gradient
  magnitude analysis
• Change Detection for Multi-band images, Large
  images, Land Cover (SAR Optical), CVA based,
  Shape in urban areas
• Hot-spot monitoring via GIS fusion Flood event
  analysis Built-up areas dynamic
• Support Functions
• Conversions from original format to Geotiff
  (Landsat, MODIS, SAR, ASAR), from Polar to
  Cartesian
• Segmentation (MEEO, ACS, SRTM / SAR, Watershed),
  Raster to Vector Converter
• GDAL Crop and Merge MERIS Band Extractor many
  elementary operators Widgets toolkit
• Parallelised Clustering Projection Pursuit

18
Multi-spectral Pre-classifier

• SOIL MAPPER (SM)Preliminary classifier
  of multi-spectral images
• Fully automated (no parameter, no data)
• Spectral rule-based (modelled from remote sensing
  knowledge in literature)
• Pixel based (very fast)
• Associating semantic meaning (40 to 80
  categories)
• Supported Missions (max no. of spectral classes)
• Landsat-like (85) Landsat TM/ETM, ASTER, MODIS,
  CBERS
• AVHRR-like (73) AVHRR 3, ATSR-2/AATSR, MSG-2
  SEVIRI
• Spot-like (59) Spot-4 HRVIR, Spot-5 HRG,
  IRS-1C/1D, IRS-P6 LISS-III/IV, AWiFS
• VHR-like (46) IKONOS, QuickBird-1, ALOS AVNIR-2

Blue / Green developed under / outside ESA
contracts
19
Landsat-7 Pre-classification
Landsat image (R band 5, G band 4, B band 1)
Classified Map
Sicily, Italy Spatial resolution 30 m
20
QuickBird Pre-classification
QuickBird image (R band 3, G band 4, B band
1),
Classified Map
Campania, Italy Spatial resolution 2.44 m
21
Area, Time Semantic Query
22
Issues for Temporal Analysis
Class. ATSR-2, Dec. 12, 1996
Large registration difference pixel-based
analysis impossible
Class. AATSR, Jan. 28, 2008
23
Possible Solution

• Re-project classified maps onto an Earth-fixed
  grid with
• 0.25 x 0.25 Tiles (Lat / Long aligned )
• 64 x 64 pixels per Tile (pixel dimension 434
  m)
• Extensible to higher resolution missions (data
  fusion, multi-sensor / multi-temporal real time
  analysis)
• Create one catalogue entry per tile (over land
  and with cloud cover lt 20 )

Class. AATSR, Jan. 28, 2008
Class. ATSR-2, Dec. 12, 1996
Earth-Fixed Grid
24
Bi-/Multi-Temporal Analysis
Graphic Evolution
Characteristics
Definition
Winter Spring Summer Fall

• Bare soil or low vegetation during winter
• Increase in vegetation during spring and early
  summer
• Bare soil from mid-summer to winter

Non-permanent crop fields (with annual cycle)
Winter Spring Summer Fall

• Bare soil during winter
• Low-to-mid vegetation during spring
• Water body on late spring - early summer
• Mid-to-strong vegetation during summer
• Bare soil during late summer/fall

Rice Fields
Low Veg.
Ice/snow
Built-up
Bare Soil
High Veg.
Water
25
Table of Contents

• Service Support
• Image Information Mining
• Reference Data Sets
• Ontology Terminology

26
Reference Data Setshttp//geonetwork.keo.esrin.es
a.int/geonetwork
27
Search Methods (1/2)
28
Search Methods (2/2)
29
Data Visibility (1/2)
Metadata might be visible to any user
30
Data Visibility (2/2)
Metadata might be visible to any user, but
downloadable only from authorised users
31
Data Download
32
Reference Data Sets (some provide links)

• (A)ATSR-like, Land Cover
• AATSR Test Dataset
• Baltic Sea Region database (BALANS)
• CORINE Land Cover 2000 (CLC2000)
• Global Climatologic-based Land Cover
  Classification (Ecoclimap)
• African Land Cover Database (FAO Africover)
• LANDSAT-like, Land Cover
• NOAA Coastal Change Analysis Program (C-CAP)
• Customized Worldwide RDS
• Minnesota Land Cover Classification System
• Regione Emilia Romagna
• Wisconsin Land Cover Use
• U.S. Landsat Based Land Cover Database (NLCD2001)
• SPOT-like, Land Use
• Bolzano land use database
• Customised Worldwide RDS
• Kalideos database
• Regione Lombardia land use database
• SPOT Test Dataset

• AVNIR-2, Cloud Cover
• AVNIR-2 Test Dataset
• Customised RDS
• Ice Applications
• Svartisen Glacier, Norway, Glacier Facies
  Classification, LANDSAT Imagery
• Hofsjokull Glacier, Iceland, Glacier Facies
  Classification, ERS/ENVISAT Imagery
• North coast of Newfoundland, Canada, Iceberg
  Monitoring, Ship Detection, ENVISAT Imagery
• Baltic Sea, Sea Ice Thickness Measurements,
  ENVISAT Imagery
• Others
• Classification, Landsat urban area points of
  interest
• MCI / FLH index extraction
• Planned
• Ortho-rectification (Maussane les Alpilles, Thun)
• Interferometry
• Co-registration (CNES, NCC)
• Change Detection
• Trend analysis

33
Table of Contents

• Service Support
• Image Information Mining
• Reference Data Sets
• Ontology Terminology

34
EO Ontology
Application Term
Multi-domain Thesaurus (shared, stable,
multi-lingual)
Semantic links
System Specific Taxonomy (dynamic, selected
language)
Application Term
Product
Service
Processor
35
Multi-domain Thesaurus
Application Terms
Multi-domain Thesaurus
36
Multi-domain Vocabulary
Multi-domain Vocabulary
37
GSCDA Shared and Specific Semantics
Multi-domain Thesaurus
Application Term
Product Type
Sensor
Mission
GSCDA Taxonomy
Dataset
GSCDA Vocabulary
Definitions
GSCDA Semantic Search Tool
Search path
Multi-domain Vocabulary
38
GSCDA Semantic Search Tool
Multi-domain Thesaurus and Vocabulary
Taxonomy and Vocabulary GSCDA
39
2D Navigator
40
2D Navigator
41
GSCDA Ontology v2
Product Type
Sensor
Mission
Dataset
Multi-domain Thesaurus
GSCDA Taxonomy
GSCDA Semantic Search Tool
Product Group
GSCDA Vocabulary
Application Term
Application Term
Definitions
Rules / Reasoning
Application Requirements
Product Group
Sensor Mode /Type
Resolution

Multi-domain Vocabulary
Band
42
Rules