AI spots dangerous cycling at intersections

A method of detecting cyclists at intersections has been developed which allows dangerous situations to be identified. Traffic monitoring cameras were used together with image processing which can automatically detect cyclists. The trajectory of the cyclists can then be tracked and potentially dangerous trajectories identified.

Series of segmented images for different types of vehicles

Image processing used to identify cyclists

In this paper, it is noted that previous research has shown that most bicycle accidents are caused by the cyclist themselves making a mistake. The most common mistakes are made during a turn (53%) and disregarding the right of way (22%), for example riding in the wrong direction against the traffic or entering a road without right of way. The objective of this research was, therefore, to identify such atypical trajectories for the movement of bicycles through an intersection. Examples include U-turns, red light violations, sudden and hard braking or cutting across the roadway diagonally. However, rather than define such situations explicitly, the researchers employed machine learning to identify atypical trajectories.

An atypical trajectory is not necessarily a dangerous one and two well-known traffic safety parameters were used to indicate the level of risk involved in atypical trajectories. The first parameter, Time to Collision (TTC) is the expected time for two objects to collide. This was determined using an Extended Kalman Filter. A low TTC indicates a dangerous situation and occurred frequently when cyclists were turning across the follow of traffic. In this case, this meant turning left since the study was carried out in Berlin with traffic traveling on the right-hand side of the road. This is known to be a common situation for an accident to occur. The second parameter, the Post-Encroachment-Time (PET), is the time by which two objects miss each other when their paths cross. It is not clear to me reading the paper why it is necessary to first identify atypical trajectories if the known parameters TTC and PET are what actually indicates danger.

It is intended by the researchers that the method will be used to analyze intersections and recommend changes to the infrastructure to improve safety.


Over the past years, the bicycle has gained importance as a means of transportation in big cities. The use and acceptance of a bicycle as being an evolving means of transportation is highly linked to its transportation safety. Still, the risk of accidents is a dominant barrier. Even though the Federal Ministry of Transport, Building and Urban Development established a National Cycling Plan to enhance cycling and improve safety aspects, serious accidents still occur. Even if the number of traffic accidents is declining in Berlin, the consequences of bicycle accidents with physical injury are characterised by increasing results. Thus, it is proved that more than half of the accidents that involve bicyclists are caused by the cyclist itself. To understand causes of accidents and to eventually arrange preventive measures and enhance cyclists’ safety, critical situations were detected. The application is based on cyclists’ trajectories generated from video sequences. As a result, atypical and dangerous traffic situations can be identified automatically whereas rule violations can be detected manually. First experiences at an intersection in Berlin show a general applicability of this approach, which has to be widely tested at other intersections.


Analysis of traffic safety for cyclists: The automatic detection of critical traffic situations for cyclists
Detzer, S. (German Aerospace Center, Institute of Transportation Systems, Traffic Management (DLR), Germany); Junghans, M.; Kozempel, K.; Saul, H. Source: WIT Transactions on the Built Environment, v 138, p 491-502, 2014, Urban Transport XX

Read the full paper…

Be the first to comment on "AI spots dangerous cycling at intersections"

Let us know what you think

This site uses Akismet to reduce spam. Learn how your comment data is processed.