Segmentation and skeleton methods for digital shape understanding

Mots clé :

• shape analysis
• discrete geometry processing
• segmentation
• skeleton

Digitised geometric shape models are essentially represented as a collection of primitives without a general coherence. Shape understanding aims at retrieving global information about the shape geometry, topology or functionality, for subsequent uses such as measurement, simulation or modification. In this context, I present my main contributions to digital shape understanding, which are mostly based on shape decomposition and skeleton computation methods. The first part explores the faithfulness of the 3D mesh representation to the real world object, through topological and perceptual analyses, and suggests a conversion to a regular volumetric model. The second part focuses on shapes in motion and details tools to create, modify and analyse temporal mesh sequences. The third part explains through two concrete examples how digital shape understanding can help experts in medicine and forestry. Finally, three open questions for shape understanding of shapes in motion and scanned trees are discussed by way of perspectives.

Raporteurs:

• Mr Adrian Hilton (Professeur - University of Surrey )
• Mme Michela Spagnuolo (Directrice de Recherche - CNR-IMATI )
• Mr Marc Daniel (Professeur - Aix-Marseille Université )

Examinateurs:

• Mr Bruno Levy (Directeur de Recherche - Inria Nancy Grand-Est )
• Mme Marie Paule Cani (Professeur - Université Grenoble Alpes )
• Mme Raphaëlle Chaine (Professeur - Université de Lyon )