Analyzing and extracting the interaction scenario between vehicles and pedestrians (V2P) are of great significance for road safety. This research proposes a method of V2P data processing and scenario extraction based on the naturalistic driving data of connected vehicles. By extracting real vehicle motion parameters, pedestrian motion parameters, and radar data from the Safety Pilot Model Deployment (SPMD), the V2P scenarios for dangerous and typical situations are extracted and analyzed. Firstly, We use the threshold method combined with manual verification to identify the real dangerous events from the data set. Then, the statistics of partial dynamic features of vehicles are chosen to be the parameters of the V2P scenarios and the Random Forest Model are used to filter out the important features as the final input of the clustering algorithm. Finally, the 4 dangerous and 6 typical V2P scenarios are extracted and analyzed. These scenarios could benefit design and testing for cooperative V2P safety applications.

CICTP