Vehicle Counting and Trajectory Detection Based on Particle Filtering

This paper proposes a new road traffic monitoring method based on image processing and particle filtering. The proposed method detects and classifies automatically moving vehicles in previously defined classes. The detected vehicles are tracked using a new particle filtering algorithm to determine their positions on the road at each time, and then the vehicle positions are used to estimate its trajectory on the road. The trajectory analysis provides information about the motion pattern (i.e. behavior) of the tracked vehicles (e.g. detect lane changes, or the most used lane on the road). Our mean contributions in this paper are the proposition of a new moving vehicle detection method, and a new tracking method based on particle filtering. The preliminary results obtained in the classification and tracking of several vehicles in five video sequences are promising. Also, these results suggest that our approach potentially allows to analyze the trajectories and identify a motion pattern for each vehicle on the road. At this stage, the results are encouraging, and shall lead to further developments in our road traffic monitoring scheme.