Edinburgh Napier University

  This paper is concerned with how video footage of pedestrian activity in various urban environments can be used to inform the development of PEDFLOW - a microscopic, agent-based model of pedestrian flow that can be used to simulate the effects of an environmental change (e.g. widening a pavement) on how pedestrians negotiate urban space. The process can be summarized as follows: (1) video footage of pedestrians is digitized and calibrated according to real-world measurements of the filming area; (2) an interesting scenario is identified (e.g. two pedestrians approaching each other on a collision course) and all cases of this scenario selected from the video for analysis; (3) for each case, qualitative information (e.g. pedestrian group size, or final action) and quantitative data (e.g. walking speed, gap size) are recorded. For the latter, the frame-by-frame position of selected pedestrians is plotted using commercially-available image analysis software and related to key features of the environment (such as kerbs and obstacles); (4) statistical analyses are performed to extract patterns of behaviour which can be translated into algorithms that form the basis of the PEDFLOW model. In this way, the model can be developed according to realistic approximations of how pedestrians negotiate urban space.