Potential and Systematisation of Video-based Analysis of Bicycle Conflicts

Abstract

Road accidents involving bicycles are becoming increasingly common in urban areas. To improve accident prevention, proactive road safety instruments (i.e., conflict analysis) are a promising supplement to the existing reactive road safety instruments. However, current research does not sufficiently account for the particularities (underestimated injury severity based on weight, speed, and collision angle) of road conflicts involving cyclists in video-based conflict analyses. To further the research on proactive safety instruments, this work developed a systematisation of existing methods using video-based conflict analysis. The recordings took place at a critical traffic node in Zurich and the resulting video data was evaluated. Conflict attributes (post-encroachment-time and time-to-collision) were calculated from the extracted trajectories. An adjusted severity formula was developed to identify and weight the potentially most relevant conflicts. The conflicts were thus analysed according to traffic mode, type of conflict, and expected collision severity, generating an analysis of conflict events at critical traffic nodes that considers the particularities of cyclists. The expected conflict severities were shown to be highest for cyclists and other vulnerable road users. These results provide an improved understanding of the conflicts and safety, supporting a more proactive approach to increasing road safety. Although the developed systematisation offers an improvement to the state-of-the-art, the collected video data has deficits. The trajectories extracted by the video remain imprecise and often cannot distinguish between the different road users. Based on this, the generation of trajectories should be further developed, and the current results of a conflict analysis should be interpreted with caution. As this work is a new attempt at an expected severity-based conflict analysis, the methodology should be further evaluated and verified at other urban traffic nodes.