PEOPLE GROUPS ANALYSIS FOR AR APPLICATIONS

Abstract

Automatically characterizing groups and crowds of people plays an important role in different domains such as psychology, architecture or entertainment. In the engineering field, people grouping is important in social signal processing but also videosurveilance. Our challenge is to introduce such solutions into the field of augmented reality, where people get added content on real groups of people.

As a preliminary work, this paper presents a system which provides a live visual feedback in virtual reality augmenting real groups of people with added information such as their ID, group ID or group coherence. The scene was analyzed with mainly one classical RGB camera and enhanced with a smartphone that a holder points towards the other persons.

This paper goes through the proposed system, which is capable of tracking people, performing people grouping, analyzing groups and augmenting those groups in a virtual world.

The first qualitative results show the feasibility of an augmented crowd environment and provide a set of interesting practical insights on the different modules of this system in the context of real-life scenes.