A Real-Time Car-Following Data Collecting Method Based on Binocular Stereo Vision

For obtaining more accurate car-following data, this study describes the design, implementation, and application of a new real-time car-following data collecting method based on a binocular stereo vision system. The car-following data consists of car-following distance, velocity, relative velocity, and driving trajectory. The system uses triangulation principle to calculate car-following distance. The system can calculate car-following distance and store the related data in real time. In the case of stable measurement, this research analyzes that the mean absolute percentage error of the distance measuring system is less than 5%. In order to solve the measurement instability caused by environmental interference, this study proposes a new state estimator based on an extended Kalman filter (EKF) technique, in which microscopic traffic flow model is used as nonlinear dynamic model. The results indicate that the proposed EKF significantly improves the data reliability in real-time car-following data collecting process.