Project

Back to overview

Multi-Camera System for Behavioral Analysis

English title Multi-Camera System for Behavioral Analysis
Applicant Fua Pascal
Number 164030
Funding scheme R'EQUIP
Research institution Laboratoire de vision par ordinateur EPFL - IC - ISIM - CVLAB
Institution of higher education EPF Lausanne - EPFL
Main discipline Information Technology
Start/End 01.01.2016 - 31.12.2016
Approved amount 86'000.00
Show all

All Disciplines (2)

Discipline
Information Technology
Sociology

Keywords (3)

Behavioral Analysis; People Tracking; Computer Vision

Lay Summary (French)

Lead
Le suivi de personnes dans des images vidéo est au cœur de nombreuses applications allant de la surveillance en environnement urbain à l'analyse comportementale dans des espaces de travail. Il a donc été une activité centrale de nos groupes de recherche respectifs depuis de nombreuses années. Jusqu'à présent, nous avons surtout opéré dans des conditions de laboratoire mais nos algorithmes sont maintenant suffisamment murs pour être déployés et testés en conditions réelles.
Lay summary

Nous allons donc installer des caméras dans le nouveau bâtiment ME de l’EPFL pour y suivre les flux de personnes,  modéliser comment le bâtiment est utilisé, et comment ces modèles d'utilisation sont affectés par les changements de mobilier. Pour éviter d'empiéter sur la vie privée des gens, nous mettrons l'accent sur les groupes et les foules et non pas sur les individus. Leur identité ne sera d’ailleurs jamais enregistrée.

L'infrastructure proposée permettra non seulement de tester nos méthodes de suivi mais aussi  de les comparer et de combiner leurs atouts. L’installation sera conçue pour être suffisamment souple pour modifier les configurations de caméra à volonté et les compléter au fur et à mesure que de nouveaux senseurs deviendront disponibles.

Direct link to Lay Summary Last update: 24.12.2015

Responsible applicant and co-applicants

Associated projects

Number Title Start Funding scheme
147693 Tracking in the Wild 01.01.2014 Sinergia
135108 Design patterns for the cross-fertilization formal and informal learning 01.02.2012 ProDoc

Abstract

Video-based people tracking is central to many applications ranging from surveillance in complex urban environments to behavioral analysis in cluttered workspaces. The research groups of the three principal investigators have been actively developing increasingly sophisticated approaches to doing this over many years. But, so far, we have been mostly restricted to testing them in a laboratory setting. Our methods are now mature enough to be deployed and tested in a real-world setting.When the new ME building is completed on the EPFL campus, the power and internet outlets needed to install the required cameras will already be in place on the ground and first floors, which makes it an ideal candidate for attempting this. There we could track the people flows, monitor how the building is used, and how these usage patterns are affected by layout changes. Of particular interest would be an in-depth study of whether or not the atrium truly functions in one the three ways it was designed to operate. Of course, to avoid invading people's privacy, we will focus on groups and crowds and not on the individuals in them. Their identity will never be recorded.The proposed image-acquisition infrastructure will not only allow for real-life tests of our tracking methods but will also make it possible to compare them and to combine their best features. The camera set-up will be designed to be flexible enough to modify the camera configurations at will and to complement them with additional imaging devices, such as depth cameras, as our understanding improves and the opportunities arise. Not only will this result in better algorithms and increased collaboration between our laboratories, it will also produce valuable data that can then be exploited by other laboratories to carry our projects in the areas listed in the proposal.
-