Tracking multiple people from standard cameras is challenging,
mostly due to the occlusions that occur as soon as several people
are involved. We tackle this problem by using several cameras,
observing the scene with different points of views. We developed a
people detection algorithm called POM that uses a
generative model of background subtraction to estimate the
positions of people in an individual time frame.
Using this
detection result, we then rely on global optimization methods,
such as Linear Programming or more recently K-Shortest Path (KSP),
to link together detections obtained at independent time
frames.
The
frame-independent people detector working together with a joint
optimization method for tracking form a very robust system that
allows to track people reliably instead of significant occlusions,
while exhibiting real-time performance.
This video shows our Linear Programming-based tracker and POM people detector working in monocular.
This video shows our recent tracking results using global apperance constraints.
Demo
We have developed
a real-time multiple tracking demo for an interactive exhibition of
the Olympic
Museum in Lausanne. The exhibition, entitled "Athletes and
Science", runs from May 2010 to March 2011, and shows the impact of
scientific developments on sport.
Our demo consists of 4 cameras
located at every corner of a small room of about 5x4m. The visitors
enter the room and are tracked in real-time by our system. Their
respective trajectory as well as the time and distance they spent
within the demonstrator are displayed to them when they exit the
room. The demo illustrates the potential application of vision-based
technologies for tracking players in team sports.
Technically, our setup is built upon
our POM people detection algorithm and our
Linear Programming-based multiple people
tracker. The system runs on a single PC and has been optimized to
attain a frame rate of 30 fps.
Source Code
The C++ source code of our POM people detector is available under a GPL license on the Software page of our web site.
The source code for our multiple object tracker is available upon request under a proprietary license for academic purposes. More information is available on the Software page of our web site.
Data Set
Some of the multi-camera video sequences that we acquired for this
project are available for download on the Data part of our web site.
Tracking Multiple People under Global Appearance Constraints
Horesh Ben Shitrit, Jérôme Berclaz, François Fleuret, Pascal Fua International Conference on Computer Vision, 2011 pdf
|
@article{BenShitrit11,
author = "H. Ben Shitrit and J. Berclaz and F. Fleuret and and P. Fua",
title = {{Tracking Multiple People under Global Appearance Constraints}},
journal = "International Conference on Computer Vision",
year = 2011,
}
Conditional Random Fields for Multi-Camera Object Detection
Gemma Roig, Xavier Boix, Horesh Ben Shitrit and Pascal Fua International Conference on Computer Vision, 2011 pdf
|
@article{Roig11,
author = "G. Roig and X. Boix and H. Ben Shitrit and P. Fua",
title = {{Conditional Random Fields for Multi-Camera Object Detection}},
journal = "International Conference on Computer Vision",
year = 2011,
}
Multiple Object Tracking using K-Shortest Paths Optimization
Jérôme Berclaz, François Fleuret, Engin Türetken, Pascal Fua IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011 pdf
|
@article{Berclaz11,
author = "J. Berclaz and F. Fleuret and E. Turetken and P. Fua",
title = {{Multiple Object Tracking using K-Shortest Paths Optimization}},
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
year = 2011,
}
Multiple Object Tracking using Flow Linear Programming
Jérôme Berclaz, François Fleuret, Pascal Fua 12th IEEE International Workshop on Performance Evaluation of Tracking and Surveillance (Winter-PETS), 2009 pdf
|
@inproceedings{Berclaz09,
author = "J. Berclaz and F. Fleuret and P. Fua",
title = {{Multiple Object Tracking using Flow Linear Programming}},
booktitle = "12th IEEE International Workshop on Performance Evaluation of Tracking and Surveillance (Winter-PETS)",
year = 2009,
month = "December",
}
Multi-Camera Tracking and Atypical Motion Detection with Behavioral Maps
Jérôme Berclaz, François Fleuret, Pascal Fua European Conference on Computer Vision, 2008 pdf
|
@incollection{Berclaz08a,
author = "J. Berclaz and F. Fleuret and P. Fua",
title = {{Multi-camera Tracking and Atypical Motion Detection with Behavioral Maps}},
booktitle = "European Conference on Computer Vision",
year = 2008,
month = "October",
pages = "112--125",
volume = "5304",
}
Multi-Camera People Tracking with a Probabilistic Occupancy Map
François Fleuret, Jérôme Berclaz, Richard Lengagne, Pascal Fua IEEE Transactions on Pattern Analysis and Machine Intelligence, 2008 pdf
|
@article{Fleuret08a,
author = "F. Fleuret and J. Berclaz and R. Lengagne and P. Fua",
title = {{Multi-Camera People Tracking with a Probabilistic Occupancy Map}},
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
year = 2008,
month = "February",
pages = "267--282",
volume = "30",
number = "2"
}
Robust People Tracking with Global Trajectory Optimization
Jérôme Berclaz, François Fleuret, Pascal Fua Proceedings of Computer Vision and Pattern Recognition 2006, New-York, USA, 2006 pdf
|
@inproceedings{Berclaz06,
author = "J. Berclaz and F. Fleuret and P. Fua",
title = {{Robust People Tracking with Global Trajectory Optimization}},
booktitle = "Computer Vision and Pattern Recognition, 2006 IEEE Computer Society Conference on",
year = 2006,
month = "June",
pages = "744--750",
volume = "1",
}
