BikeDetect
Contact
BikeDetect
BikeDetect - Development and testing of a cost-efficient sensor system for improved detection of cyclists in real road traffic
Cyclists are often exposed to dangerous situations in road traffic. The minimum overtaking distance defined in the StVO (1.5 m in urban areas) is often not observed by motorists, partly because this cannot be seen with certainty by “visual inspection”. There is little knowledge about critical situations with insufficient side clearance when overtaking or passing cyclists from the perspective of motorized traffic. Based on an improved data situation, motorists could adapt their driving behavior and thus indirectly contribute to more attractive and safer cycling.
The aim of “BikeDetect” is to develop a sensor system in the form of a laboratory model for retrofitting motorized traffic in order to be able to detect distances to cyclists in road traffic to increase cycling safety. The optimum sensor system is being designed on the basis of several pre- and field tests. The data evaluation combines distance measurements with the automated detection of cyclists in road traffic using AI methods. Practical safety requirements of municipal planning (e.g. minimum distances, execution speed of the AI models) are incorporated into the implementation.
The requirements for the sensor system and AI models (e.g. permissible error rates, GDPR) are determined in a workshop. Laboratory tests investigate combinations of sensors for distance (ultrasound, optics, radar) and classification (thermal, LiDAR, 3D camera) in terms of measurement accuracy and robustness. In the field test, the prioritized sensor system will be tested on selected routes in the city of Osnabrück using larger vehicles (vans). The data generated will be evaluated using data science methods and visualized on the basis of municipal requirements using newly selected key figures and maps on road safety.