Quasi-Random Sampling for Condensation

The problem of tracking pedestrians from a moving car is a challenging one. The Condensation tracking algorithm is appealing for its generality and potential for real-time implementation. However, the conventional Condensation tracker is known to have difficulty with high-dimensional state spaces and unknown motion models. This paper presents an improved algorithm that addresses these problems by using a simplified motion model, and employing quasi-Monte Carlo techniques to efficiently sample the resulting tracking problem in the high-dimensional state space. For N sample points, these techniques achieve sampling errors of O(N-1), as opposed to O(N-1/2) for conventional Monte Carlo techniques. We illustrate the algorithm by tracking objects in both synthetic and real sequences, and show that it achieves reliable tracking and significant speed-ups over conventional Monte Carlo techniques.

[1]  L FoxBennett,et al.  Implementation and tests of low-discrepancy sequences , 1992 .

[2]  Michael Isard,et al.  Contour Tracking by Stochastic Propagation of Conditional Density , 1996, ECCV.

[3]  Russel E. Caflisch,et al.  Quasi-Random Sequences and Their Discrepancies , 1994, SIAM J. Sci. Comput..

[4]  P. Fearnhead,et al.  An improved particle filter for non-linear problems , 1999 .

[5]  Timothy F. Cootes,et al.  Building and using flexible models incorporating grey-level information , 1993, 1993 (4th) International Conference on Computer Vision.

[6]  Harald Niederreiter,et al.  Programs to generate Niederreiter's low-discrepancy sequences , 1994, TOMS.

[7]  Paul Bratley,et al.  Algorithm 659: Implementing Sobol's quasirandom sequence generator , 1988, TOMS.

[8]  Andrew Blake,et al.  Classification of human body motion , 1999, Proceedings of the Seventh IEEE International Conference on Computer Vision.

[9]  Andrew Blake,et al.  Learning Dynamics of Complex Motions from Image Sequences , 1996, ECCV.

[10]  Bennett L. Fox,et al.  Algorithm 647: Implementation and Relative Efficiency of Quasirandom Sequence Generators , 1986, TOMS.

[11]  Tomaso A. Poggio,et al.  Pedestrian detection using wavelet templates , 1997, Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[12]  William H. Press,et al.  Numerical Recipes: The Art of Scientific Computing with IBM PC or Macintosh , 1996 .

[13]  Michael J. Black,et al.  Recognizing temporal trajectories using the condensation algorithm , 1998, Proceedings Third IEEE International Conference on Automatic Face and Gesture Recognition.

[14]  Michael Isard,et al.  ICONDENSATION: Unifying Low-Level and High-Level Tracking in a Stochastic Framework , 1998, ECCV.

[15]  Harald Niederreiter,et al.  Implementation and tests of low-discrepancy sequences , 1992, TOMC.

[16]  J. E. Glynn,et al.  Numerical Recipes: The Art of Scientific Computing , 1989 .

[17]  R. Caflisch,et al.  Smoothness and dimension reduction in Quasi-Monte Carlo methods , 1996 .

[18]  Michael Isard,et al.  Learning Multi-Class Dynamics , 1998, NIPS.

[19]  P. Fearnhead,et al.  Improved particle filter for nonlinear problems , 1999 .

[20]  William H. Press,et al.  Numerical recipes in C. The art of scientific computing , 1987 .

[21]  Harald Niederreiter,et al.  Random number generation and Quasi-Monte Carlo methods , 1992, CBMS-NSF regional conference series in applied mathematics.

[22]  Dariu Gavrila,et al.  Real-time object detection for "smart" vehicles , 1999, Proceedings of the Seventh IEEE International Conference on Computer Vision.

[23]  Andrew Blake,et al.  Learning dynamical models using expectation-maximisation , 1998, Sixth International Conference on Computer Vision (IEEE Cat. No.98CH36271).

[24]  David C. Hogg,et al.  Learning Flexible Models from Image Sequences , 1994, ECCV.

[25]  Les A. Piegl,et al.  The NURBS Book , 1995, Monographs in Visual Communication.

[26]  Dariu M Gavrila,et al.  REAL-TIME OBJECT DETECTION USING DISTANCE TRANSFORMS , 1998 .