Reclustering for Large Plasticity in Clustered Shape Matching

Author/Creator ORCID



Type of Work


Computer Science and Electrical Engineering


Computer Science

Citation of Original Publication


This item may be protected under Title 17 of the U.S. Copyright Law. It is made available by UMBC for non-commercial research and education. For permission to publish or reproduce, please see or contact Special Collections at speccoll(at)
Distribution Rights granted to UMBC by the author.


In this theses, we present a novel contribution to the clustered shape matching framework. Clustered shape matching describes an algorithm introduced a decade ago by M�ller and colleagues in 2005, which was designed to allow for deformable bodies to behave in a physically plausible way when it comes to collisions, deformations, or fractures. This is accomplished by sampling the given deformable object with particles and clustering the particles together in such a way that they can accurately reflect plastic and elastic deformations in the object. These sampled particles determine the degrees of freedom present within the object. At each timestep, a best-fit rigid transformation of the rest of the shape of the object to the current configuration of particles is computed, and Hookean springs are used to pull the particles towards the transformed shape. Clustered shape matching algorithms of this nature have proven to be robust enough to offer realistic physical simulations, while also being efficient enough to run in real-time, offering a level of interactivity for the user. One limitation of clustered shape matching becomes apparent during large plastic deformations. Recently, there was a piece of research work published that extended basic clustered shape matching in an attempt to address this limitation by dynamically adding and removing clusters and particles. In this theses, we re-visit this limitation and propose a more careful, principle-driven solution to the problem of reclustering. Additionally, we show through experimentation that our proposed solution does not change the behavior of the material of the sampled object. Furthermore, we demonstrate that the particle reclustering is sufficient in our framework to handle extremely large plastic deformations, allowing us to easily conserve the mass of the object. Lastly, we present a concrete example, highlighting an error in estimating rotations in the original shape matching work of M�ller and colleagues that has persisted for over a decade through other followup work in shape matching.