The population grows daily and with it the number of tourists heading for popular spots, attractions and cities. Research in the International Journal of Security and Networks has considered one aspect of the safety of large crowds, the sheer number of people that might be present in a given location. Qinqin Dong of Xinyang University, China, points out that the biggest challenge in managing dense groups of people in real-time is determining how many people are present in a crowd.
Dong has turned to artificial intelligence to develop a new took that can track and trace the movements of people in a crowd with unprecedented accuracy. The new system, SMACSTR (Scene Monitoring Algorithm based on Crowd Scene Type Recognition), could allow us to improve safety in bustling urban spaces and popular destinations.
The behaviour of a crowd is largely unpredictable unless barriers and other measures are in place to guide their movements. At popular tourist spots, there can be many hundreds or even thousands of people moving in unexpected surges that can represent a risk to safety. Overcrowding leads to bottlenecks and if an emergency arises, hazardous crowd movements that need to be addressed and controlled quickly to avoid injury and death.
Dong’s SMACSTR system can carry out crowd scene recognition, to interpret images of crowds and their behaviour. The system focuses on both static and dynamic features within the crowd and allows its operators to spot risky behaviour as it arises and so be able to implement a timely and effective response.
The static component of the system, the static density field, reveals crowd positioning and numbers, while the dynamic, the motion feature maps, indicate how the crowd is behaving. It can distinguish between calm and peaceful movements of individuals in the crowd or the emergence of erratic or panicked behaviour. By combining both characteristics of the crowd, the system can assess risks more effectively in real-time in a way that conventional systems, such as human monitoring of CCTV feeds, might not.
Dong, Q. (2024) ‘Safety monitoring system for tourist scenic spots based on crowd scene type recognition’, Int. J. Security and Networks, Vol. 19, No. 3, pp.128–137.