Deformation Capture and Modeling of Soft Objects

Bin Wang, Longhua Wu, Kangkang Yin, Uri Ascher, Libin Liu, Hui Huang

We present a data-driven method for the deformation capture and physics-based modeling of soft deformable objects. Our framework enables both realistic motion reconstruction and synthesis of virtual soft object models in response to user stimulation. Low cost depth sensors are used for the deformation capture, and we do not insist on any force-displacement measurements which are commonly required by previous work, thus making the capturing a cheap and convenient process. A physics-based probabilistic tracking method is employed to increase the tracking robustness to noise, occlusions, fast movements and large deformations. The deformation estimation task that includes the reference shape, material elasticity parameters and damping coefficient is then formulated and solved as a spacetime optimization problem, aiming at matching the simulated trajectories with the tracked ones. The optimized deformation parameters are used in turn to make the physics-based tracking results more accurate, consequently improving the deformation estimation itself. Numerical experiments demonstrate that our physics-based deformation tracking and deformation parameter optimization can be unified and made complementary to each other. The obtained optimal deformation parameters can yield high quality animations for various soft models.

Deformation Capture and Modeling of Soft Objects