How to Measure Safety Risks for Cyclists at Intersections?

More and more people perceive the bicycle as an attractive mobility alternative. To confirm even more people for a more frequent bicycle use a safe infrastructure, especially at intersections where most of the severe bicycle accidents are happening, is a key prerequisite. The problem here is that knowledge about cycling safety is often limited to accident statistics which are reported to the police. Therefore, information about near accidents or even near interactions among cyclists is not available. To tackle this problem, the Province of Antwerp (Belgium) tested a 3D camera system by the Swedish company Viscando to detect near interactions at a dangerous intersection in the municipality of Bornem (Province of Antwerp) in September 2019. This data driven implementation is part of the research project BITS (Bicycles and Intelligent Transport Systems) which is funded by Interreg B as part of the North Sea Region (NSR) programme. As the project is focused on the topic of open bicycle data and its potential to increase safety, speed and comfort for cyclists, new questions arise about data quality, how to publish the near interaction data set as open data and what kind of new knowledge about safety of cyclists at intersections could be generated as part of the data analysis. The collected database was processed and further analysed by the University of Oldenburg, Department of Business Informatics (Very Large Business Applications VLBA), Germany. The first problem which had to be solved was that the camera data was not only provided without any images but also without any geoinformation. The total amount of 114 conflict points contains x and y coordinates on a locally limited coordinate system which needed to be transformed into a longitude and latitude format. In a next step the speed levels of the cyclists were calculated to learn more about potential risks. The intersection can be divided into areas with higher and lower speed levels. By dividing the near interactions into different times of the day it can be learned that the frequency and the average speed levels of the cyclists are increasing in the morning hours. A higher risk to be involved in a near interaction when many people cycling to work or school could be concluded. In a last step the cycling directions of the cyclists involved in the near interactions were determined. These seem to be very relevant to learn more about the degree of risk of a certain near interaction. Contrasting directions in combination with high (or differing) speed levels could increase the probability of an accident. Based on the findings it will be discussed how to measure risks for cyclists at intersections in general.