Recognizing safety-critical events from naturalistic driving data

Sammanfattning: New trends in research on traffic accidents involve conducting Naturalistic DrivingStudies (NDS). NDS are based on large-scale data collection of driver, vehicle andenvironment information in real-traffic. NDS provide large data sets which have provento be extremely valuable for the analysis of safety-critical events such as near crashesand incidents.NDS data needs to be filtered to recognize safety-critical events. Filtering safety-criticalevents has been traditionally achieved by using kinematics triggers (e.g. searching fordeceleration below a certain threshold signifying harsh braking). The low sensitivity andspecificity of this filtering procedure, however, requires manual annotation of video datato decide whether the events individuated by the triggers are actually safety-critical.Such reviewing procedure is based on subjective decisions, time-consuming, and oftentedious for the analysts.This project looked into improving this reviewing procedure using video data collectedfrom 100 Volvo cars during one year in Gothenburg within a NDS called euroFOT.More than 400 videos from the triggered events have been reviewed, concluding thatdriver’s reaction may be the key to discriminate safety-critical events. In fact, whetheran event if safety-critical or not depends on the driver. Several statistical procedureshave been then applied to automatically recognize driver reaction from video data. Inthis project, we showed how combining automated video analysis with kinematicstriggers increases sensitivity of near crash recognition from NDS data. These resultsopen up to new ways to use video frames in NDS.