3D Geometry and Animation
Permanent members – 2 ETPR: Dominique Bechmann (Prof.), David Cazier (Prof.), Franck Hetroy-Wheeler (Prof.), Pierre Kraemer (Associate Prof.) Engineers: Joris Ravaglia, Sylvain Thery Post-doc: Paul Viville (ANR Posture project: 2023-2025) PhD student: Qijia Huang (12-22 to 2025)
Over the last few years, the team has gradually developed skills in volumic modeling in which 3D objects are represented by an internal decomposition into volume cells and not only by their surface. These representations are particularly important for many methods of solving equations simulating physical phenomena (such as fluid flows) with applications in fields such as health, urban planning, and engineering. We intend to continue the exploitation of our uniform topological models in several dimensions in order to propose effective original methods for the generation of volumetric meshes adapted to specific use cases by developing collaborations with local teams of fluid mechanics. Temporal animation and the dynamic adaptation of meshes are avenues of development that we intend to pursue. Among the fields of application targeted, we can cite, for example, the simulation of particle trajectories in the bronchi during breathing cycles or the simulation of flow in a model of a moving heart.
Numerical processing of geometry
In this theme, the team also highlights geometry in a broader way than geometric modeling alone. Point clouds constitute the vast majority of data produced by current 3D digitization technologies and form a common base for several projects (already initiated and to come) of this theme. In particular, the members of the theme wish to continue their research on the analysis of the geometry of plants from digitizations. Indeed, the geometry as the topology of plants being extremely variable even within a given species, these plants constitute a particularly difficult object of study, for which classical methods (generally suitable for manufactured objects) fail. We propose to direct our research around several axes such as the filtering and the registration of point clouds, their segmentation and their classification, the temporal follow-up of an evolving object, or the completion of occultations in point clouds. . These axes aim to extract information from digitized objects. Neural networks specialized around point clouds have reached sufficient maturity in recent years but sometimes remain limited. This is why we will rely on a combination of “classical” processing, based on purely geometric tools, and deep learning. The study of plant geometry also guides future collaborations on the theme. The theme then offers more possibilities for interdisciplinary interactions, in particular with colleagues in agronomy, forestry, ecology or fundamental biology, and will benefit from more opportunities.
The UrTrees project proposes developing a citizen science approach to facilitate data collection on urban trees. Indeed, an inventory of urban trees and their characteristics is essential to identify better the ecosystemic services they provide and to better understand their functioning and their resilience to climate change. The project includes three successive stages. The first stage consists of developing a citizen science application on smartphones allowing the collection of data by any user. These data will be exploited in the second stage via photogrammetry techniques to derive key indicators at the tree level. The third stage aims to use these enriched data to model the services provided by trees in an urban environment. UrTrees is a collaborative project with geographers (LIVE laboratory, University of Strasbourg) and ecologists (The Open University, United Kingdom) funded by the MITI of the CNRS.