KEI Project

Project Title   KEO Extensions and Installations
Project Acronym   KEI
Contractor(s)   ACS, UTV, MEEO


Image Context
       How it will Work



Since the year 2000 the European Space Agency has started a series of activities which aimed at implementing Image Information Mining (IIM) techniques for improving the analysis of huge amounts of Earth Observation satellite data. These efforts led during years to design the KEO (Knowledge-centred Earth Observation) prototype system, a modular and scalable Component-based Processing Environment which permits to:

  Image Ease the access to EO data and relevant information extracted from them
  Image Provide a large set of tools for EO data processing (bridging the gap between Data and Information)
  Image Expand the use of EO data by supporting and automating the identification and extraction of information relevant for users
  Image Encourage the use of a common scientific cooperative environment

The KEO prototype permits users to interactively extract relevant features and information from EO data, either through a generic probabilistic technique (KIM subsystem) or by means of specific processing algorithms (FEP subsystem), and to provide outputs, i.e. valuable information extracted from data, in easily accessible formats. Since larger and larger quantities of higher and higher resolution EO images are acquired from an increasing variety of sensors and stored in archives reaching or surpassing the petabyte size, emerging big applications (e.g.: change detection, global monitoring, disaster management support, etc.) will require more and more information to be extracted. More automated, direct and human centred methods should be provided for the information extraction process, which should rely on "intelligent" (knowledge-based and learning) and easy to use (semantic driven interactions) programming environments.




The general purpose of the KEI project is to carry out a number of activities to enhance Image Information Mining (IIM) capabilities provided by the KEO prototype and to foster its use at ESRIN and partners premises. In particular, the key tasks are:

  Image Analysis and evaluation of KIM capabilities against other methods
  Image Extension of KIM functions to cope with new requirements (e.g.: add new capabilities)
  Image Installation of the KEO prototype at partners’ premises for validation in specific applications
  Image Extensions of KEO functions to cope with new requirements (e.g.: new Processing Components)
  Image Exploitation in KEO of Spectral Categorisation techniques

The project (kicked-off on November 2006) is organized as a frame contract over three years and it is carried out by a consortium, led by ACS together with UniversitĂ  Tor Vergata (UTV) and MEEO (Meteorological and Environmental Earth Observation), which guarantees .




The architecture of the KEO prototype is shown in the figure below, where system main components are represented:

  Image The KAOS Client Application, which permit to access to all system functionalities
  Image The KIM subsystem, based on the Probabilistic Information Mining (PIM) theoretical concept for interactive analysis and identification of image features
  Image The FEP subsystem, which permits to create and execute user-defined processing chains for extracting information from images
  Image OGC Web Servers, where the extracted information can be stored for reuse


The KEO environment provides users with the following functionalities:

  Image Create & semantically identify internal / external Processing Components
  Image Create Processing Components from KIM training (also interactive use)
  Image Graphically chain Processing Components into more complex Processing Chains
  Image Store output into Web Servers (WFS, WMS, WCS)
  Image Create and publish Web Services (from Processing Chains or output)



How it will Works

Under the KEI project different tasks are planned to be carried out, in order to evaluate and improve the KEO prototype:

KIM performance evaluation

KIM classification and objects detection performance will be quantitatively evaluated, considering different cases (e.g. alga bloom detection from MERIS, Landsat classification, etc.) and comparing with other algorithms and methodologies reported in literature. A specific comparison between KIM classification and Neural Network classification will be also considered.

New Feature Extraction Algorithms in KEO

New feature extraction algorithms will be selected and implemented. Among them, particular attention will be dedicated to neural networks algorithms. The use of neural networks in remote sensing has often been found effective. With respect to classical methods, neural networks represent a fundamentally different approach to classification problems, as they do not rely on probabilistic assumptions and do not need particular requirements about normality in data sets. Moreover, they show a considerable ease in using multi-domain data sources, since they can simultaneously handle nonlinear mapping of a multidimensional input space onto the output one.

KEO Reference Data Sets implementation 

New KEO Reference Data Sets will be implemented, collecting different types of images and sets of ground truth on pixel base for specific test sites. Published articles, reports and studies over the same sites will be also included.

KEO Processing Components and Processing Chains for classification

New KEO Processing Components will be developed for providing classification tools (based on the Image Spectral Categorization algorithm) for Landsat and SPOT data, possibly extending them to a large variety of satellite sensors. Related SSE services will be provided as well.

Deployment of KEO and MIMS prototypes system onto new hardware at ESRIN

This activity will consist in the migration of both KEO and MIMS systems to the new hardware platform procured by ESA. The KEO prototype will be also upgraded with the new functionalities developed and implemented for the analysis of time series during the IIM-TS project.

KEO and KIM development

The activity will include the reconfiguration of KIM Ingestion Chain into an open architecture, permitting to plug-in new primitive feature extraction modules, as well as the integration of a series of new primitive feature extractors to improve system capabilities (Gray Level Co-occurrence Matrix, Gabor Texture Descriptors, Pauli and Freeman decomposition, etc.)

New Scientific Collaborative Environment

The activity consists of setting-up a scientific collaborative environment to maximize the fruits of interactions between scientific partners working on common research fields.

Neural Network Toolbox Enabling the Generation of Customized Components

The activity foresees the development and integration into KEO of a user-friendly software toolbox for image classification based on Neural Network techniques. The toolbox will permit to design new classification algorithms and to upload them into the KEO system a new Processing Components.

New Processing Components for Hyperspectral Imagery

Development and integration into the KEO system of new processing to exploit hyperspectral data for road infrastructure monitoring, implementing image processors for both higher (such as MIVIS) and lower (such as PROBA) spatial resolutions.

New Processing Components and Application Models for land use management

The task will cover a range of activities which aim at enlarging KEO system capabilities. New proposed themes are listed hereafter:
- Analysis and implementation of Data Management and Application Models
- Implementation of environmental monitoring applications based on RS data for vegetation assessment, burned areas identification, snow cover mapping and inland water bodies assessment
- Estimation of carbon stocks




In the frame of the currently defined tasks, the KEI project will provide basic modules and automatic procedures for image processing to enrich the KEO environment, as well as related new SSE services.



Contributors to this page:

Page last modified on Tuesday 28 of December 2010 16:26:47 CET by .