Multi-target Data Association by Tracklets with Unsupervised Parameter Estimation

Abstract

We consider multi-target tracking via probabilistic data association among tracklets (trajectory fragments), a mid-level representation that provides good spatio-temporal context for making correct data association decisions efficiently. A novel approach is presented to infer the optimal model parameters and the trajectories for a varying number of targets in an unsupervised manner. The parameter estimation and the search for the best association among tracklets are unified naturally within a Markov Chain Monte Carlo sampling procedure. Experimental results demonstrate that the algorithm is able to adapt to different video sequences with occlusion and background clutter.

Citation

Paper thumbnail Weina Ge and Robert T. Collins. Multi-target Data Association by Tracklets with Unsupervised Parameter Estimation , 19th British Machine Vision Conference (BMVC), September, 2008. [PDF]

Poster

Results

Acknowledgements and Funding

This research is supported by: