1 Crowd Monitoring: State-of-the-art and Future Directions Utkarsh Singh, Jean-François Determe, François Horlin, Philippe De Doncker U. Singh was with the OPERA – Wireless Communications Group at Université libre de Bruxelles. He is now with the Artificial Intelligence Lab, Vrije Universiteit Brussel, Brussels 1050, Belgium. (Corresponding author, E-mail: utkarsh.singh@ulb.ac.be) J.-F. Determe is with the OPERA – Wireless Communications Group at Université libre de Bruxelles, Brussels 1050, Belgium. (E-mail: jde- terme@ulb.ac.be) F. Horlin is with the OPERA – Wireless Communications Group at Université libre de Bruxelles, Brussels 1050, Belgium. (E-mail: fhorlin@ulb.ac.be) Philippe De Doncker is with the OPERA – Wireless Communications Group at Université libre de Bruxelles, Brussels 1050, Belgium. (E-mail: pdedonck@ulb.ac.be) Abstract— With the growing concerns over public safety, the importance of crowd monitoring is being realized by various security and event management agencies worldwide. Estimation of crowd dynamics can help such agencies in prevention of any unanticipated accidents or issues. Research on crowd monitoring has been underway since the past few decades. Con- ventional crowd monitoring systems mainly rely on computer vision approach. Due to predominant use of videos/ image sequences, the existing techniques may raise data privacy concerns. This has led to development of new crowd monitoring techniques which are privacy preserving and require minimum public participation. This paper aims to serve as a single and sufficient source of information to the concerned researchers on various aspects of crowd monitoring and also provide future directions which can be helpful in developing advanced crowd monitoring techniques. Keywords—Artificial intelligence, Crowd monitoring, Computer vision, Detection, Forecasting, Sensor Networks, Public safety 1. INTRODUCTION Crowding is a common phenomenon observed during major events such as concerts, festivals, sports, games, and entertainment. There is a potential risk to public safety during such large scale gatherings. Recent examples include the mishaps during 2006 Mecca pilgrimage in Saudi Arabia and 2010 Love Parade in Germany [1]. Crowd monitoring helps in estimation of crowd dynamics; which can help in better event management and ensuring public safety. It is therefore, imperative for the event organizers to monitor the crowd, detect/predict possible risks and take possible mitigative measures. Besides, crowd monitoring has various added advantages, such as: • It can help in better layout of the venue of event. For example, well-designed food points, shopping points, entries, and exits can prevent overcrowding during an event. • Analysis of crowd behaviour can help in preparing crowd management strategies in overcrowded events such as concerts, Olympics, and demonstrations. • When used at a micro-level (e.g. facial recognition), it can help in tracking and apprehending suspects. • It can be useful in developing intelligent technologies for automated crowd assistance in large-scale events. • Crowd monitoring can also give ideas for virtual simulation and testing of crowd behaviour. The early research has mainly focused on crowd monitoring systems using image or video sequences (vision based approach) [2]. However, there is limited possibility of using such systems for real-time situational awareness [3]. Efforts have also been made to identify and define parameters associated with crowd dynamics. Parameters like crowd density, velocity, and flow direction can be crucial in risk analysis [4]-[6]. With the technological advancements, the research is being gradually redirected from vision based approach towards sensor based crowd monitoring systems. Modern technologies like Smartphone, GPS and Bluetooth have already been proposed for estimation of crowd location, density, and dynamics during major events [7]- [10]. However, even the recent crowd monitoring 1 2 2 3 6 2 1 3 3 2 6 6 7 13 22 37 54 54 75 67 79 90 124 149 173 28 0 20 40 60 80 100 120 140 160 180 200 1995 2000 2005 2010 2015 2020 Dcouments Year Fig. 1. Documents published per year on ‘crowd monitoring’. [Source: Scopus]