Loading...
 
ESA > Join & Share > Technology Projects > CARD Project
Print

CARD Project

Project Title   Classification Application-services and Reference Datasets
Project Acronym   CARD
Contractor(s)   MEEO (Italy)

 

Project Context
          Objectives
             Architecture
          How it works
Output

 


Context

Land cover classification is one of the most important Remote Sensing (RS) applications. The knowledge of the biophysical cover of the Earth's surface represents one of the main information source for being aware about the ecosystem health and transformation (e.g.: deforestation, urbanisation, desertification, etc...), both at global and local scale. In the past several years, many different techniques have been developed. Differences between each other depend from the approach (a priori or a posteriori), the classification scheme (hierarchical or non hierarchical), the scale of monitored area (global or local), the level of human interaction (supervised or unsupervised) and other parameters which are specific aspects of the particular application.

Due to the huge quantity of available data, the (semi-)automation of such procedures is becoming more and more important. This is the direction currently followed by ESA within the Knowledge-centred Earth Observation (KEO), KIM Extension and Installations (KEI) and Image Information Mining - Time Series (IIM-TS) projects. From the results of these automations ESA aims at obtaining Web Services to be published in the Service Support Environment (SSE).

The SOIL MAPPER® software, property of the MEEO company, leads in the same direction, since it permits automatic pre-classification of images based on spectral signatures.

In this context, the Classification Application-services and References Datasets (CARD) Project aims at reviewing existing methodologies for image information extraction, implement state of the art algorithms and generate, from some of these implemented, SSE services to be accessed through the SSE portal, providing at the same time reference datasets to validate elaboration results.

Top  


Objectives

The objectives of the CARD project can be summarized as follows:

  Image Analyse possible new methodologies for automated extraction of information from one image or image time series, also based on the use of the SOIL MAPPER® software;
  Image Make use of the SOMAFID fire detection system and extend and complement the SOIL MAPPER® to create, from MODIS images, new SSE services related to fires and burned areas detection;
  Image Define, create and link to KEO two RDSs in support to land cover and land use classification activities, building from the material collected for testing the SOIL MAPPER® software;
  Image Set-up and provide, for promotional use, services based on the SOIL MAPPER® software;
  Image Implement a complete processing system and SSE services for AATSR data based on the SOIL MAPPER® and SOMAFID software.


Main innovation of the CARD project resides on the automatic nature of each of the implemented modules, that permits to develop basic "cellular" processor for modular complex services realization.

A further innovation involves the stratified nature of each developed module: the possibility to work on a multi-layer stratified semantic basis permits to specialize the second level elaboration modules on the feature to be analysed avoiding any background pixel contamination.

Top

 


Architecture

The CARD project is organized as follows:

  1. Analysis and implementation of new methodologies (prototype software tools) for:
    1. Detection of Burned Areas:
      • Fire detection in a single image;
      • Burned area identification:
        • Potential burned area identification in a single image;
        • Burned area identification in a bi-temporal image pair.

    2. Biomass Evaluation:
      • vegetation indexes;
      • vegetation change detection techniques.

    3. Single image detection Algorithms:
      • Study of single image processing systems for image information extraction like
        • Radiometric calibration;
        • Spectral classification;
        • Feature extraction systems (value added products like spectral indexes and masks);
        • Morphological filters.
      • Second level stratified image processing systems like:
        • Grass/Forest/Urban area detection;
        • Morphological filter application (top hat of opening and top hat of closing);
        • Ship detection;
        • Stratified image segmentation
        • Topographic correction.

    4. Time series detection Algorithms:
      • bi/multi-temporal semantic-based change detection;
      • (absolute) difference between a bi-temporal map pair;
      • automatic dark subtraction (bias estimation);
      • automatic relative calibration;
      • stratified image matching;
      • unsupervised change detection from value added products.

  2. Implementation of four SSE Services for:
    • Fire Detection by single image;
    • Burned area identification by bi-temporal image pair;
    • Vegetation change assessment by bi-temporal image pair.
  3. Definition and implementation in the KEO environment of the following reference datasets for land use and land coverage validation studies:
    1. Landsat Land Cover Reference Data Set;
    2. SPOT Land cover change/land use Reference data Set;
    3. AATSR Land Classification Reference Data Set
  1. Realization of a series of SSE services based on the SOIL MAPPER® software for Landsat, SPOT, AVHRR and MODIS data open to the community until June 30st 2010;
  1. Configuration of a dedicated HW/SW system for the analysis of AATSR data; in particular the following processing systems have been implemented:
    • a series of SSE services based on the SOIL MAPPER® and SOMAFID software modules for AATSR data;
    • a complete image processing chain to classify the entire ATSR-2 and AATSR archives;
    • a SSE Service to access via spatial, temporal and semantic queries to the ATSR2-AATSR classification maps archive.

Top


How it works

  The CARD Project evolution followed four main phases:

Phase 1. Analysis of possible new methodologies for automated extraction of information from one image or image time series and implementation of basic image processing software modules;

The output from this phase is a series of technical notes that surveys the state of the art and proposed new methodologies for:

  • fire and burned areas detection;
  • biomass estimation;
  • single image processing applications;
  • time series processing applications.

The following software modules have been provided as prototype software modules for specific applications:

  • Vegetation assessment
    • Vegetation change assessment from bi-temporal images;
  • Single image enhancement modules:
    • Topographic Correction
    • Second stage processing module for Grass/Forest/Urban area detection
    • Ship detection
  • Bi-temporal image enhancement modules:
    • Relative radiometric calibration.
    • Stratified histogram matching

Phase 2. Engineering of specific software modules and implementation of SSE Services for fire and burned areas identification.
The following SSE Services have been implemented and made available to the community by June 30 2010:

Services description, user access modes and user manual can be found on the Service Level Agreement File.


Phase 3. Generation of KEO-based Reference datasets for land use / land cover validation
The following RDS have been implemented in Geonetwork.


Each RDS refers to a specific application / sensor. Together with ground truth and documentation, a set of test images of the reference sensor is also provided, in compliance with the provided ground truth to support the validation process (both input images and ground truth information are available for a specific sensor, oever a specific area in a specific time frame).

The access to the Reference Datasets is restricted to authorised users. Access to the reference datasets can be requested sending an e-mail to Andrea Della Vecchia (ESA/ESRIN).

Phase 4. Implementation and application of a dedicated HW/SW structure for AATSR data processing.
A specific hardware system that permits to process the entire ATSR-2 / AATSR was procured, configured, installed and set to operation. The system ingested all the archived data through a Tape interface, and downloads automatically from the rolling archive any new AATSR image received.
All pre-classified maps have been stored in a  dedicated geographic, temporal and semantic database.
A specific SSE Service has been implemented to access to the database in real time

The ATSR2-AATSR Classification Map Data Catalogue service permits to perform geographic, temporal and semantic queries and retrieve all the archived data in real time

Horizontal Activities
Horizontally, a series of SSE services based on the SOIL MAPPER® software for Landsat, SPOT, AVHRR and MODIS data have been made available to the community.

Following services are available for user-provided images supporting the following formats: CEOS Landsat 5, CEOS Landsat 7, Landsat 7 Geocover, Landsat 7 Level 1G, DIMAP Spot 4, DIMAP Spot 5, HDF MODIS, ENVISAT (A)ATSR, AVHRR/3-SHARP3, AVHRR/3-KLM

More details about these services can be found on the Service Level Agreement Document.

Following services are available for Category 1 users in general. Available input images can be selected  from the ESA Landsat and ESA Modis Databases:

More details about these services can be found on the Service Level Agreement Document.

Following services are available for SPOT Category 1 users in general. Available input images can be selected  from the ESA SPOT Database:

More details about these services can be found on the Service Level Agreement Document.

Services based on user-provided images remains available free of charge until June 30 2010. Services restricted do Category 1 users do not have any time limitation.

Top

 


 

 

Output

The project outputs can be classified as follows:

  • technical documentation (see Phase 1 above);
  • prototype software modules (see Phase 1 above);
  • SSE Services (see Phase 2, Phase 4 and horizontal activities above);
  • Reference Datasets implemented on Geonetwork (see Phase 3 above);
  • (A)ATSR processing system (see Phase 4 above).

 


Contributors to this page: .

Page last modified on Wednesday 22 of December 2010 15:08:56 CET by .