This study evaluated the accuracy of truck data collected by dual-loop detectors on Seattle area freeways. The objectives of the study were to 1) quantitatively evaluate the accuracy of a representative sample of dual-loop measurements of vehicle volumes and vehicle classifications on the Seattle FLOW system, 2) identify the types and causes of dual-loop data inaccuracies, and 3) recommend appropriate methods for improving the quality of real-time dual-loop measurements.
One representative station with four dual-loop detectors was selected for an extended error analysis. The detector measurements were compared to ground truth data collected via a video image system. Data for 20-second intervals were compared for peak and off-peak periods. Since there was no documentation on the settings of the dual-loop system clock, an independent analysis was conducted to synchronize the dual-loop and video image systems clocks before analysis.
In general, three dual-loop error types were identified through comparison of dual-loop and video ground-truth data: (1) underestimation of vehicle volumes, (2) incorrect assignment of Bin 3 vehicles to Bin 4, (3) incorrect assignment of Bin 2 vehicles to Bins 1 and 3.
Dual-loop measurement errors such as those described may be due to defects in system hardware, software, or the underlying measurement algorithm. Because hardware errors were virtually eliminated as a factor in the current study, the serious errors that still occurred were most likely due to defects in the underlying dual-loop algorithm or in the implementation program. Therefore, to radically improve the quality of dual-loop data, and thus the quality of real-time truck data on the FLOW system, the Washington State Department of Transportation dual-loop algorithm and its corresponding implementation code should be the emphasis of future research.
Washington State Transportation Center (TRAC)
Accuracy, Algorithms, Data collection, Data quality, Error analysis, Loop detectors, Software, Traffic volume, Truck traffic, Vehicle classification, Video imaging detectors.