Visual perception-based identification of pedestrians' close contact behaviors

In public places, tracking the movement trajectories of key personnel under investigation, and their close contacts tracing is an important task in public health and public security domains. This paper proposes a method for identifying close contact relationships between targets using visual perceptual information. First, based on the spatiotemporal characteristics of target trajectories, 15 spatio-temporal topological relationships between dynamic targets and 5 spatio-temporal topological relationships between dynamic and static targets were defined. Next, using the spatio-temporal trajectory information of the moving targets extracted from the video and taking the spatial topological relationship between the targets and their coexistence time as indicators, five pedestrian association states are classified: chance encounter, stay, walking together, close contact and co-occurrence. Finally, experiments conducted in an indoor scenario demonstrate the effectiveness of the proposed method. This research offers a novel solution for personnel tracking in public safety and public health, and plays a significant role in advancing the theory and methods for microscopic-scale crowd dynamics observation.