Automated Road Safety Analysis using Video Sensors
by Nicolas Saunier
The importance of reducing the social and economic costs associated with traffic
collisions can not be overstated. The main goal of this research is to develop a
method for automated road safety analysis using video sensors in order to
address the problem of deteriorating historical collision data. The method will
automate the extraction of traffic conflicts (near misses) from video sensor
data. There are two main research directions. The first is based on the learning
and prediction of vehicle movements. We worked on the clustering of trajectories
using hidden Markov models (HMMs) and the identification of conflicting
clusters. The second approach consists in classifying conflict and non-conflict
interactions. For that purpose, we train an ensemble of HMMs on misclassified
instances. In both approaches, experiments on limited real world data show
promising results.
Back to the LCI Forum page