LIP6 1999/016

  • Thesis
    Modélisation spatio-temporelle floue pour la reconnaissance d'activités militaires
  • V. Eude
  • 296 pages - 12/17/1998- document en - http://www.lip6.fr/lip6/reports/1999/lip6.1999.016.ps.gz - 4,297 Ko
  • Contact : Virginie.Eude (at) nulllip6.fr
  • Ancien Thème : APA
  • This search concerns military intelligence. The problem is to recognise the global activity of the enemy with the information perceived on the theatre of operations. "Typical" activities of the enemy known by experts are described with templates, hierarchical trees giving the decomposition of global activities into sub-activities. These activities are linked with different kind of constraints, among which temporal and spatial constraints, the most important ones.
    These constraints are imprecise and we model them with fuzzy sets. Each template corresponds to many fuzzy constraints graphs, one for temporal constraints and three others for spatial constraints of positions, distances and directions. The reasoning is based on the propagation of constraints and the calculation of the minimal graph, corresponding to the optimal propagation. For spatial constraints we just have an approximation of the minimal graphs.
    Fuzzy constraints allow the refinement of the evaluation of the template's assumption, because they may be partially satisfied. A template is generally partially recognised, some sub-activities have not been observed. The global evaluation of templates must take into account the satisfaction of the constraints, the activities recognised and their interpretation in the decomposition.
    We introduce some satisfiability measures for the first part of evaluation and different aggregation operators with their interpretation in our context for the second part. We extend symmetric summations to g-symmetric summations.
  • Keywords : Temporal reasoning, spatial reasoning, fuzzy constraints, CSP, FCSP, aggregation operators
  • Publisher : Valerie.Mangin (at) nulllip6.fr