Probabilistic Deformable Surface Tracking from Multiple Videos: A
Patch-Based Approach
Slobodan Ilic
(TU Munich, Germany)
Abstract:
We address the problem of tracking the temporal evolution of
arbitrary shapes using multi-camera information. This is motivated by the
growing number of applications that require consistent shape information
along temporal sequence for various purposes such as motion analyzes. The
approach we propose considers a temporal sequence of independently
reconstructed surfaces and iteratively deforms a reference mesh to fit
these observations. In contrast to methods using strong prior models, this
framework assumes little on the observed surface and hence easily
generalizes to most free-form surfaces. To this aim, the reference surface
is divided into elementary surface cells or patches. This strategy ensures
robustness by providing natural integration domains over the surface,
while enabling to express simple patch-level rigidity constraints.
process, such as fake volumes and missing parts. Motion is often fast and
surfaces undergo large deformations. To effectively cope with such data,
the problem is cast as Bayesian maximum-likelihood estimation where the
joint probability of the deformation parameters, i.e motion, and the
observed data is to be maximized. The problem is solved using EM algorithm,
where in the maximization steps we introduce a generic numerical
optimization that solves for physically plausible surface deformations
given arbitrary constraints. This combination of a generic mesh
deformation approach with uncertainty models allows for evolutions of
arbitrary meshes driven by noisy observations. It successfully handles
missing data, relatively large reconstruction artifacts, fast motion, large
deformations and multiple objects.
Extensive experiments demonstrate the effectiveness and robustness of the
method on various standard and novel 4D datasets.
Short Bio:
Since February 2009 Slobodan Ilic is a leader of the Computer
Vision Group of the CAMP Laboratorie at TUM, Germany. Form June 2006 he was
a senior researcher at Deutsche Telekom Laboratories in Berlin. Before that
he was a postdoctoral fellow for one year at CVLAB, EPFL, Switzerland,
where he received his PhD in 2005 under supervision of Prof. Pascal Fua.
His research interests include: deformable surface modeling and tracking,
3D reconstruction, real-time object detection and tracking, object
detection and localization in 3D data. Slobodan Ilic serves as a regular
program committee member for all major computer vision conferences, such as
CVPR, ICCV and ECCV as well as journals, such as TPAMI and IJCV and is an
Area Chair of ICCV 2011. Besides active academic involvement Slobodan has
strong relations to industry and has a number of students supported by
industry, such as Deutsche Telekom, BMW, MVTec, ESA and Siemens.