Traffic Monitoring of Motorcycles During Special Events Using Video Detection

Because of a recent federal initiative, states are required to submit motorcycle vehicle miles traveled data to the FHWA. Data are needed to obtain better counts of motorcycles to evaluate their impact on crashes and traffic flow, but there is concern about the quality of data. Many states have identified problems with automatic traffic recorders accounting for motorcycle traffic. Existing sensors exhibit difficulties counting motorcycles that travel side by side or close behind each other, they have difficulty distinguishing larger motorcycles from passenger vehicles, and magnetic counters in particular do not sense motorcycles that do not pass over or travel close enough to sensors. Alternatively, some states conduct manual classification counts, which are labor intensive and lead to sparse data. Because classification counts are frequently conducted during the week, they do not capture weekend numbers. This paper evaluates a video-based traffic monitoring system that was developed at Clemson University and can classify vehicles, including motorcycles. The processor uses vehicle tracking, rather than virtual detection, to collect vehicle count, speed, and classification data. An algorithm calculates a motorcycle's (l)ength, width, and height through a series of frames. The system is evaluated by using traffic data for more than 2,000 motorcycles, collected at two locations in Myrtle Beach, South Carolina, during a motorcycle rally. The difference between actual and system motorcycle counts ranged from 0.6% to just over 6%, depending on direction and location. The difference for all vehicles ranged from 0.25% to 3.6%. While the system successfully classifies motorcycles traveling in close pairs and in small groups, it experienced difficulty in cases of severe occlusion.