This paper presents an algorithm to classify vehicles passing by a detector into two categories:long vehicles (LVs) and passenger cars (PCs), using high-resolution event-based loop detectordata (every vehicle-detector actuation is recognized as an event). By long vehicles, we meanvehicles whose lengths are at least 2 or 3 times that of ordinary passenger cars (about 19 feet long).Since vehicle speed and vehicle length are closely related, one cannot estimate the vehicle speedwithout knowing the vehicle length, or vice versa. However, when a group of vehicles forming aplatoon, their relative speeds are rather deterministic based on Newell's simplified car-followingmodel. Since the time gaps between consecutive vehicles and the detector occupation time canbe easily derived from event-based data, vehicle platoons can be identified with the time gaps anda least square minimization program can be formulated to estimate the speed of the first vehiclein the platoon and the platoon acceleration rate, using the measured and the estimated detectoroccupation time. The vehicle classification algorithm is tested using the event-based detector datacollected from Trunk Highway 55 in Minnesota and the estimation results are verified using theconcurrent video.
展开▼
