Exploring Variability in Cyclists Riding Posture through Naturalistic Data: A Computer Vision and Bayesian Modelling Approach

Övrigt konferensbidrag, 2024

Vulnerable Road Users (VRUs) are frequently involved in road traffic accidents, accounting for more than half of all road traffic deaths (WHO, 2023). Among these VRUs, cyclists, who are exposed to a high risk of injuries (Cittadini et al., 2024; Stigson et al., 2020), form a significant portion. The severity and type of injuries sustained by cyclists can be influenced by various factors, including riding posture, helmet usage, bike type, and speed. Previous research, such as the study conducted by (Leo et al., 2023), has explored the posture and riding preferences of VRUs (specifically e-scooter riders) in a controlled field experiment. Controlled experiments, while valuable, have their limitations. To overcome these limitations, this study utilizes naturalistic data, which offers insights into rider behaviors and postures under real-world conditions. Video data collected from cyclists in a naturalistic setting combined with computer vision algorithms can provide information not only about their posture but also about their sex and helmet usage. By understanding and modeling the posture of cyclists, we can leverage tools such as human body models and simulations to recreate various crash scenarios. The recreated crashes allow us to gain a deeper understanding of injury mechanisms, ultimately contributing to decreasing cyclists’ injuries.