Among the algorithms developed towards the goal of robust and efjcient tracking, two approaches which stand out due to their success are those based on particle filtering [8, 12, 141 and variational approaches [S, 161. The Bayesian approach led to the development of the particlejilter, which peforms a random search guided by a stochastic motion model. On the other hand, localising an object can be based on minimising a cost function. This minimiin1 cun be found using variational methods. The search paradigms differ in these two methods. One is stochastic and modeldriven while the other is deterministic and data-driven. This paper presents ci new algorithm to incorporate the strengths of both approaches into one consistent framework. To allow this fusion a smooth, wide likelihood fiinctiorl is constructed, based on a sum-of-squares distance measure and an appropriate sampling scheme is introduced. Based on low-level information this scheme automatically mixes the two methods of search and adapts the computational demands of the algorithm to the di’culty of the problem at hand. The ability to effectively track complex motions without the need forfinely tuned motion models is demonstrated.
[1]
N. Gordon,et al.
Novel approach to nonlinear/non-Gaussian Bayesian state estimation
,
1993
.
[2]
Michael J. Black,et al.
A Probabilistic Framework for Matching Temporal Trajectories: CONDENSATION-Based Recognition of Gestures and Expressions
,
1998,
ECCV.
[3]
Dorin Comaniciu,et al.
Real-time tracking of non-rigid objects using mean shift
,
2000,
Proceedings IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2000 (Cat. No.PR00662).
[4]
Michael Isard,et al.
Partitioned Sampling, Articulated Objects, and Interface-Quality Hand Tracking
,
2000,
ECCV.
[5]
David J. Fleet,et al.
Stochastic Tracking of 3D Human Figures Using 2D Image Motion
,
2000,
ECCV.