Field Testing for Automated Identification of Turning Movements at Signalized Intersections

Obtaining turning-movement counts at signalized intersections is a routine task in traffic engineering and can be tedious and time-consuming. Previous research in automating turning-movements counts focused on estimating the turning movements from approach and departure volumes and developing detection systems for exclusive-turn lanes. The accuracy of an alternative method, called the time-and-place system (TAPS), is examined through a field study of five signalized intersections in Columbia, Missouri. TAPS uses both the locations and times of actuations from a small number of detectors to classify movements from shared approach lanes. The five intersections are representative of different geometrics and signal timings. At four intersections a standard video camera was placed at a height of about 30 ft and as close to the departure lanes as possible to provide a reasonable view. Additional cameras showed current signal indications into the departure leg. At the fifth location a single elevated camera captured both vehicle movements and signal indications. The videotape data were used to compare TAPS results with actual flows. The errors in detection were apparently due to the sensitivity of the detection system, camera angles, intersection geometrics, traffic parameters, and other factors. The ability of TAPS to identify turning movements at signalized intersections was supported by the study results. With ideal camera views the authors believe that TAPS could yield almost perfect results. The information from TAPS could enhance the capabilities of real-time traffic adaptive signal control systems, dynamic traffic assignment, and traffic demand estimation.