An approach to detect left-turn forbidden intersection using taxi trajectories

Recently, the trajectory mining of moving objects is becoming the focus of researchers. In the aspect of ITS (Intelligent Transport Systems), floating cars equipped with GPS produce trajectories continuously so long as they are traveling on the road. These large amounts of trajectories imply extensive and valuable traffic information. In this paper, we mining from a large number of real-world taxi trajectories and propose an approach to detect the intersections with left-turn forbidden constraint. We build our system based on a trajectory dataset produced by about 15,000 taxis during a period of one month, and evaluate the system by comparing with the real-world data came from in-the-field test. The assessment results show that the correct rate of our method can be up to more than 75%, which as is believed, will have great benefit on the manufacture of precise navigation equipment.