Avoiding the "streetlight effect": tracking by exploring likelihood modes

Classic methods for Bayesian inference effectively constrain search to lie within regions of significant probability of the temporal prior. This is efficient with an accurate dynamics model, but otherwise is prone to ignore significant peaks in the true posterior. A more accurate posterior estimate can be obtained by explicitly finding modes of the likelihood function and combining them with a weak temporal prior. In our approach, modes are found using efficient example-based matching followed by local refinement to find peaks and estimate peak bandwidth. By reweighting these peaks according to the temporal prior we obtain an estimate of the full posterior model. We show comparative results on real and synthetic images in a high degree of freedom articulated tracking task.

[1]  Trevor Darrell,et al.  Inferring 3D structure with a statistical image-based shape model , 2003, Proceedings Ninth IEEE International Conference on Computer Vision.

[2]  Paul J. Besl,et al.  A Method for Registration of 3-D Shapes , 1992, IEEE Trans. Pattern Anal. Mach. Intell..

[3]  Nebojsa Jojic,et al.  Tracking self-occluding articulated objects in dense disparity maps , 1999, Proceedings of the Seventh IEEE International Conference on Computer Vision.

[4]  Michael Isard,et al.  Attractive People: Assembling Loose-Limbed Models using Non-parametric Belief Propagation , 2003, NIPS.

[5]  David J. Fleet,et al.  Stochastic Tracking of 3 D Human Figures Using 2 D Image Motion , 2000 .

[6]  Olivier D. Faugeras,et al.  3D articulated models and multi-view tracking with silhouettes , 1999, Proceedings of the Seventh IEEE International Conference on Computer Vision.

[7]  David J. Fleet,et al.  People tracking using hybrid Monte Carlo filtering , 2001, Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001.

[8]  David J. Fleet,et al.  Stochastic Tracking of 3D Human Figures Using 2D Image Motion , 2000, ECCV.

[9]  Cristian Sminchisescu,et al.  Estimating Articulated Human Motion with Covariance Scaled Sampling , 2003, Int. J. Robotics Res..

[10]  Stan Sclaroff,et al.  Database Indexing Methods for 3D Hand Pose Estimation , 2003, Gesture Workshop.

[11]  Ankur Agarwal,et al.  Learning to track 3D human motion from silhouettes , 2004, ICML.

[12]  Trevor Darrell,et al.  Constraining human body tracking , 2003, Proceedings Ninth IEEE International Conference on Computer Vision.

[13]  K. Rohr.  Towards model-based recognition of human movements in image sequences , 1994 .

[14]  Nebojsa Jojic,et al.  Tracking articulated objects in dense disparity maps , 1999 .

[15]  R. Plankers,et al.  Articulated soft objects for video-based body modeling , 2001, Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001.

[16]  Andrew Blake,et al.  Articulated body motion capture by annealed particle filtering , 2000, Proceedings IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2000 (Cat. No.PR00662).

[17]  Michael Isard,et al.  CONDENSATION—Conditional Density Propagation for Visual Tracking , 1998, International Journal of Computer Vision.

[18]  Trevor Darrell,et al.  Fast pose estimation with parameter-sensitive hashing , 2003, Proceedings Ninth IEEE International Conference on Computer Vision.

[19]  Trevor Darrell,et al.  Bayesian Articulated Tracking Using Single Frame Pose Sampling , 2003 .

[20]  James M. Rehg,et al.  A multiple hypothesis approach to figure tracking , 1999, Proceedings. 1999 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Cat. No PR00149).

[21]  Piotr Indyk,et al.  Similarity Search in High Dimensions via Hashing , 1999, VLDB.