A study on roaming behaviour of crowd in public space with the analysis in computer vision and Agent-based simulation

  • YEAR
    Wu, Yuansheng
    Zhang, Zipei
    Daii, Shimada
    Liu, Changlin
    Niu, Shaoyu
    Xia, Lipeng
    Xiao, Liangfa
    Ikeda, Yasushi
    2022 Conference Papers
    Conference Papers


People-flow in densely populated modern cities is a non-negligible factor to consider during urban space design. However, in the early phase of design, architects mainly deal with the static states of the human body. Other factors, such as duration and environment, should be considered when analysing crowd behaviour. The common methods for design presentation are drawings, sketches, models, etc. Dynamic visualisation of pedestrian behaviour might help architects have a better understanding of the design performance. Agent-based simulation has been explored by many researchers. Most of their pedestrian models have planned routes with origin and destination. Thus, we would like to propose a pedestrian model embedded with roaming behaviour that reflects the decision-changing process when exploring unfamiliar places. Previously, computer vision and agent-based simulation were different research streams. This paper discusses the initiative to study crowd behaviour using computer vision and agent-based simulation and examined the accuracy and validity of agent-based simulation by comparing the results from computer vision. Overall, this research aims to improve the accuracy of the agent-based simulation by utilizing the data extracted from surveillance video.

Keywords: Agent-based Simulation; Computer Vision; Crowd Behaviour; Data Visualisation.


To top