Plan du site Contact Calendrier Nouveau Home

Banque de données projets FEDRA


Actions de recherche



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

Updating rate assessment system of topo-geographical data using space remote sensing (ETATS)

Projet de recherche S0/00/021 (Action de recherche S0)

Personnes :

  • Dr.  BRUYNSEELS Hugues - Institut Géographique National (IGN)
    Partenaire financé belge
    Durée: 2/1/2003-30/9/2005
  • M.  ACHEROY Marc - Ecole Royale Militaire (ERM)
    Partenaire financé belge
    Durée: 2/1/2003-30/9/2005

Description :

Context and objectives

NGI is currently completing the entire coverage of the Belgian territory with topo-geographical data at the scale 1:10 000, taking this opportunity to re-think its overall process workflow. Up to now, the topo-geographical data are managed in different scale-related databases. Separated updating processes for each scale are performed by systematic field survey and no information exists about the actualisation status of the data. NGI is setting up a project to reorganise all the data in one seamless database, to reduce the data production means and to define a common strategy to update the data and automatically generalise them to other scales. The main goal of the ETATS project is to provide the decision makers with general information about large changes relative to the road network and the built up areas. This information (presented as GIS layer or map), coupled with other data coming from the cadastre, the Flemish, Walloon and Brussels regions, etc., will provide a basis for human analysis and planning of the updating process. The ETATS objective is then to design and implement a tool for automatically detect the important changes of the road network and of the built up areas comparing the NGI database to recent earth observation data (SPOT5).


Changes are located by comparing a mask generated by the NGI data base (DB) projected on the image, to the output of a classifier extracting the built-up area and the road network, called the "man-made'' or MM class. NGI filters the DB to produce vector layers containing only the built-up area, the road network, and the hydrography. The first two layers are used to produce the ”Old Mask'', representing the old extent of the MM class. The SPOT5 images are then registered with the vectorial DB. In parallel, on the one hand, the registered panchromatic image is analyzed by a "Texture and Structure algorithm'' that separates textured from non textured areas, as at this scale, built-up areas and roads generate texture. On the other hand, the NDVI (Normalized Difference Vegetation Index) computed from the multi-spectral images, provides another two-classes separation: vegetation and non vegetation areas. The fusion of both classifications from which the hydrological network is removed, is compared with the "Old mask'' to generate a "Change Map''. Note that, in the final report, we recommend not to use the NDVI because of its response over bare soils.


The main result of the project is an application able to automatically produce a change map of the road network and the built up areas comparing the NGI database to SPOT5 images. The resulting map is made of a layer of regions in red, yellow and green, corresponding respectively to a disparition, an apparition or a status quo of the objects of interest. The application is subdivided into several modules (global georeferencing, fine registration with vector data, vector to raster transformation of the NGI database, NDVI computation, change detection, and finally export to shape file and tif raster files). This subdivision non only allows the end user to segment his work but also to use other specific procedures or already processed data that could be available from other sources. The application comes with an end-user manual and a set of recommendations. Another result of the project is a program computing a quality index based on the morphological gradient. An index below 6 and above 10 qualifies a poor and a good quality image respectively. The highest the quality is, the better the change detection results will be.

Documentation :

A propos de ce site

Cookies policy

Données personnelles

© 2020 SPP Politique scientifique