The Hybrid Model for Lane-Changing Detection at Freeway Off-Ramps Using Naturalistic Driving Trajectories

In order to promote traffic safety at freeway off-ramps, this paper designed a hybrid model to identify a lane-changing with vision technology. The unmanned aerial vehicle was used to collect video stream data at five off-ramps for Xi'an Raocheng freeway during weekdays. The positional information-of-an individual vehicle is recorded at a frequency of 5 Hz. Each trajectory is composed of 30 positional records and all trajectories are divided into lane-changing and lane-keeping units. Features such as lateral driving speed, lane departure, and the lane deviation angle extracted from trajectory records are related to the lane-changing behaviors. We develop a hybrid model of the Gaussian Mixture Model and the Continuous Hidden Markov Model to identify lane-changing behaviors at off-ramps with these features. Basing on test set, we conduct a test for the hybrid model and the result shows that the prediction accuracy of the proposed model is as high as 94.4% for lane-changing behavior and 93.6% for lane-keeping behavior.