Plan du site Contact Calendrier Nouveau Home

Banque de données projets FEDRA

Presentation

Actions de recherche

Personnes

Chercher

Recherche et applications > Banque de données projets > Banque de données projets FEDRA

Development of a spatio-temporal segmentation algorithm for satellite time series to monitor forest condition (ECOSEG)

Projet de recherche SR/01/108 (Action de recherche SR)


Personnes :

  • Prof. dr.  COPPIN Pol - Katholieke Universiteit Leuven (K.U.Leuven)
    Coordinateur du projet
    Partenaire financé belge
    Durée: 1/12/2006-31/5/2008
  • Dr.  CULVENOR Darius - Commonwealth Scientific and Industrial Research Organisation (CSIRO)
    Partenaire financé étranger
    Durée: 1/12/2006-31/5/2008

Description :

Context and objectives

Various studies have explored the use of satellite time series to describe the seasonal dynamics of vegetation. Typically, the approaches do not take into account the spatial or hierarchical context of the data. Hence, the aim of ECOSEG is to improve the integration of temporal information into hierarchical image segmentation.

ECOSEG a spin-off project of GLOVEG VG/00/01 and encompasses three objectives: (i) improvement of existing multi-temporal hierarchical image segmentation (MTHIS) via the development of a new methodology (EMTHIS) to hierarchically cluster image time series data into similar spatio-temporal segments at a range of scales, (ii) implementation, validation and optimization of EMTHIS on remote sensing datasets and (iii) dissemination of EMTHIS tool to the international scientific community.

The present proposal is submitted by a consortium of the Geomatics research unit of the Katholieke Universiteit Leuven (KULeuven) and Ensis – a joint venture between Australia’s Commonwealth Scientific and Industrial Research Organisation of and the New Zealand Forest Research Institute Limited (CSIRO-Scion).


Methodology

A novel conceptual methodology (EMTHIS) is introduced to improve the existing MTHIS. This allows the hierarchical clustering of image time series into spatio-temporal segments with similar time series at various scales. The principle of the proposed methodology is analogous to the classical hierarchical image segmentations which use bottom-up region-merging techniques but the innovative feature of EMTHIS is its definition of initial objects. EMTHIS handles a pixel with its temporal properties over a specified time window as the initial object, instead of considering a pixel with its spectral (HIS) or entire temporal properties (MTHIS). The development of EMTHIS will initially be conducted using computer modeling simulations to develop the method and efficient algorithms, and conduct accuracy assessment. After validation of the method on synthetic data, the EMTHIS will be applied, validated and optimized on actual satellite time series data to ensure the utility of the technique in later research and to national and international partners. Remote sensing time series of biophysical indicators of forest health (e.g. defoliation and discoloration) will be employed as key indicators in this context.


Results expected

At the end of this 18 months project, a well-defined user-friendly stand alone tool with graphical user interface (GUI) accompanied by a detailed user manual will allow a non-specialist to hierarchical cluster image time series into spatio-temporal segments with similar time series at numerous scales and will allow an accuracy assessment based on common accuracy statistics. Moreover, we will deliver proof of the importance of the integration of temporal information into hierarchical image segmentation and we will demonstrate the added-value of EMTHIS where it concerns the study of satellite time series of ecosystem dynamics. Based on the experimental setup, an extensive study of the EMTHIS methodology will deliver a detailed insight in the pros and cons for a variety of applications.
Unique is also the fact that we will have brought together complementary experience to guarantee a state of the art study of multi-temporal hierarchical image segmentation, hopefully having established a permanent “knowledge center” for multi-temporal hierarchical image segmentation.


Documentation :

A propos de ce site

Déclaration de confidentialité

© 2017 SPP Politique scientifique