Real-Time Identification of Vehicles on Highways by 3D Model Matching under Stop-and-Go Conditions

Robust vehicle reidentification, gives important information about the level of service (LOS) on highways and freeways. Apart from being useful in off-line decision-support tools for design and planning, vehicle reidentification also gives real-time feedbacks about: traffic conditions, mean flow rate, lane occupancy, and vehicle classification. Such information improves the driver awareness and safety, and increases the efficiency in dealing with jams and other traffic-perturbing events. Recently many vehicle reidentification systems have been proposed, that show appropriate behaviour when dealing with freeflow traffic conditions. Unfortunately, they show unacceptable performance when dealing with stop-and-go traffic conditions. In this paper we present a robust, real-time, laser-based, vehicle reidentification system, which uses 3D volumetric signatures and does not make assumptions about the speed of the vehicles. So the system gives realiable and accurate measurements even during stop-and-go traffic conditions. The proposed architecture uses an AUTOSENSE 600 laser camera, coupled with a TV color camera. The color camera estimates displacement information in real-time, that is used to correct the measurements obtained by the laser camera. Simulations give an overall correct reidentification rate of about 0.96 . Nowaday the system is under test on real working portals; the preliminary experimental data are in good agreement with the simulations