Thèse de doctorat en Informatique
Soutenue le 12-12-2012
à Dijon , dans le cadre de École doctorale Sciences pour l'ingénieur et microtechniques (Besançon ; Dijon ; Belfort) , en partenariat avec Laboratoire Electronique, Informatique et Image (LE2i) (Dijon, Côte d'Or ; Auxerre, Yonne ; Chalon-sur-Saône, Saône-et-Loire ; Le Creusot, Saône-et-Loire) (laboratoire) .
Le président du jury était Klaus Böhm.
Le jury était composé de Joerg Klonowski, Andreas Nüchter.
Les rapporteurs étaient André Borrmann, Gilles Gesquière.
Déduction des relations spatiales qualitatives à partir d'opérateurs géométriques quantitatifs : une nouvelle approche appliquée à la détection et la qualification sémantique d'objets 3D
Pas de résumé
This work presents the 3D Spatial Qualification tool (3DSQ) which was created tocompute spatial data stored in OWL-DL ontology. By using the adjustment principle ofan existing ontology, it is then possible to add 3D data to existing objects and computetheir spatial relationships from their 3D models. The 3DSQ Platform makes an attempt toensure the interaction between heterogeneous environments. Actually, such a semanticplatform connects an adjusted OWL ontology structure, a 3D quantification engine, avisualization engine and a set of geometry via knowledge processing technologymaterialized via SWRL, SQWRL rules and SPARQL queries within its extended Built-Ins. The created Spatial Built-Ins are connected to the presented quantification engine andenable qualifying semantic spatial relationships. This will mainly help us to not just applysemantic queries selecting geometry based on such a qualified relationship, but also tobenefit from the richness of the knowledge based schema, from a logical point of view. Itincludes the semantic definition and the implementation of the standard 3D spatialrelationships and uses sophisticated geometry data structure like NEF Polyhedra. Itfurther describes the implementation of the suggested bridge by the means of the NEFPolyhedra operation and the DLs definition of spatial relation.In addition, this thesis presents an application of the 3DSQ platform. It is argued that therepresentation of spatial information is not a fundamental limitation of OWL, wherelinking top level semantic qualification with low level quantitative calculation is highlypossible and efficient via the OWL-DL expressive power. This efficiency is carried outby the semantic rule system, and the geometry data structure required for therepresentation of spatial regions. In fact, such a semantic qualification based ondescription logic (DLs), and OWL ontologies enable much more efficient and intelligentspatial analysis semantically. To prove the feasibility and to validate the 3DSQ Platformwithin its quantitative and qualitative 3D spatial operators, real applied areas related toBuilding Information Model (BIM), IFC and especially 3D point clouds data wereaddressed. Given the complexity of the underlying problems, the suggested new methodsresort to using semantic knowledge, in particular, to support the object detection andqualification. In this context, a novel approach which makes use of the 3DSQ platformand benefits from intelligent knowledge management strategies to qualify objects will bediscussed. It is based on the semantics of different associated domains to assist inknowledge formalization where Knowledge helps in the qualification process, and can beclearly palpable through the thesis.Such a conception will bring solutions to the problem raised by the syntactic exchangelevel between CAD software packages, IFCs or 3D point cloud geometries. Moreover, allrelations between the different geometries are defined by elements suggested in thisthesis. In fact, these relations define how elements can interact. Such a semantic can onlybe synthetized, used and invested by OWL ontology structure with all the robustness ofthe Description Logics