Real Time Vehicle Trajectory Estimation on Multiple Lanes

Today’s road traffic management systems using intelligent transportation systems solutions need real time measurements of various traffic parameters like flow, origin-destination matrices, vehicle type, etc. Cameras combined with image processing algorithms are being more and more used as the sensor capable to measure several traffic parameters. One such parameter, also important for accurate simulation of road traffic flow and evaluation of traffic safety, is the driving aggressiveness factor which can be estimated from the vehicles trajectory. In this paper an Extended Kalman Filter based approach to estimate vehicle trajectories on multiple lanes using only one static camera is described. To test the accuracy of the implemented approach a synthetic road traffic environment is developed. Real time capabilities of the approach are tested using real traffic video footage obtained from Croatian highways.