MCMC-Based Feature-Guided Particle Filtering for Tracking Moving Objects from a Moving Platform
Abstract
This paper proposes a Markov Chain Monte Carlo based feature-guided particle filtering algorithm to track moving objects observed from a camera on a moving platform. Sudden camera or object motion is the typical problem that causes tracking performance sharply deteriorate. It is inadequate to use classical recursive Bayesian estimation to track moving objects observed by a rapid-moving and unstable camera since the method could not resolve the sudden motion problem. We develop a robust and unconstrained tracking algorithm to overcome the tracking failure issues. Markov Chain Monte Carlo (MCMC) technique is adopted to efficiently realize the feature-guided particle filter. Experiment results show that the method demonstrates robust tracking performance without assistance of foreground segmentation and performs accurately in severe tracking environment.
Cite
Text
Lin and Wolf. "MCMC-Based Feature-Guided Particle Filtering for Tracking Moving Objects from a Moving Platform." IEEE/CVF International Conference on Computer Vision Workshops, 2009. doi:10.1109/ICCVW.2009.5457616Markdown
[Lin and Wolf. "MCMC-Based Feature-Guided Particle Filtering for Tracking Moving Objects from a Moving Platform." IEEE/CVF International Conference on Computer Vision Workshops, 2009.](https://mlanthology.org/iccvw/2009/lin2009iccvw-mcmcbased/) doi:10.1109/ICCVW.2009.5457616BibTeX
@inproceedings{lin2009iccvw-mcmcbased,
title = {{MCMC-Based Feature-Guided Particle Filtering for Tracking Moving Objects from a Moving Platform}},
author = {Lin, Chung-Ching and Wolf, Wayne H.},
booktitle = {IEEE/CVF International Conference on Computer Vision Workshops},
year = {2009},
pages = {828-833},
doi = {10.1109/ICCVW.2009.5457616},
url = {https://mlanthology.org/iccvw/2009/lin2009iccvw-mcmcbased/}
}