Research action on Automation of generalisation


Continuing the research action GNOTO, GIGA works on automation of the generalisation process, in taking into account secondary objects, more complex than before, in studying the non-cartographic generalisation and in adding some mechanisms of process introspection to facilitate the development of the system.

The action research in generalisation has been active from 2005 to 2009

Leader of the research action :
Cécile Duchêne



The COGIT laboratory has been studying generalisation processes with cartographic goal since 1992. The approach adopted by the COGIT is the local step to step approach (cf. location of the research topic outside IGN). This approach consists in locally applying algorithms to the data in response to locally identified conflicts. A conflict corresponds to a non-respect of the specifications: over-density of objects, too-detailed geometry, local superposition... The research carried out until 2001 has lead to a better characterisation of information (PhDs of Corinne Plazanet, Pascale Monier, Jean-François Hangouët, Annabelle Boffet) on the design of new algorithms (Emmanuel Fritsch, Nicolas Regnauld), on the strategies of generalisation (Anne Ruas, Cécile Duchêne), on the consolidation of knowledge for automation (Sebastien Mustière)and as well, on the evaluation (Sylvain Bard). The PhD of Frederic Hubert, undertook within the research project on the data consultation and attended in July 2003, contributes to the help of the capture of the generalisation specifications(user needs).

These researches made advances possible on 3 points :

  • the implementation of generalisation algorithms more relevant to objects nature,
  • the implementation of measurements enabling to identify conflicts and to characterise space in order to better choose the algorithms and evaluate the results,
  • the implementation of 2 complementary generalisation models containing agents which enable to control the sequencing, application space and to choose the algorithms.


Objectives :

  • at short-term, we should couple togethers AGENT and CartACom models, so as to take the advantages of both models.
  • cartographic generalisation of themes, as land cover or forms of the terrain for the topographic data.
  • improvment of procedural knowledge used by a generalisation system.
  • generalisation of databases, i.e. generalisation with a non-cartographic goal

COGIT PhD on this topic


IGN Web Site