Pedestrian Crossing Style Recognition Based on Multi-source Parameter Fusion

Pedestrians are the vulnerable group in the urban traffic system. It is of great significance to explore the characteristics of pedestrian crossing behavior to reduce human-vehicle conflict and improve the safety of pedestrian crossing. This paper designed a pedestrian crossing experiment under two scenarios (free crossing and psychological critical crossing) to investigate the coupling relationship among crossing speed, psychological critical time window, visual characteristics, and other major parameters. The risk perception factor and critical safety factor were selected as the representation indexes of pedestrian style under psychological critical crossing state. The crossing style recognition model was developed to analyze the reliability of different styles of pedestrian crossing behavior. The results show that the age of pedestrian significantly affects crossing speed/mental critical time window. The psychological critical time window is negatively correlated with crossing speed/signal light area fixation probability and positively correlated with traffic area fixation probability. In the state of psychological critical crossing, from the result of style recognition, male is more cautious than female. The failure rate of risky crossing behavior is 65.22%. There are significant deficiencies in the assessment ability of moving function and visual exploration features of some elderly pedestrians performing risky crossings. The study results provide reference for the design and optimization of intersection facilities, pedestrian crossing education and training, etc. © 2022 Science Press. All rights reserved.