The location of the stop-bar is the prerequisite for traffic parameter estimation and advanced traffic control at urban intersections. Existing literature has explored to identify intersection layout through various methods, such as trajectory data mining and remote sensing. However, few studies have addressed the identification of stop-bar location (IoSBL). At signalized intersections, stop-and-go phenomena are commonly observed due to periodical green time allocation. With the assumption that the IoSBL is equivalent to the positioning of the first vehicle in a queue, this paper proposes a statistical approach to identify the location of the first vehicle in a queue by exploiting the stop-and-go phenomena based on trajectory data. Firstly, the phenomena are captured when vehicle speeds are low and the variance is large. Secondly, a spatial-distribution density function of the phenomena is drawn. Thirdly, the location of the stop-bar is identified by applying a density percentile threshold (DPT). The positions below the DPT are considered invalid. For the remaining part, the position nearest to the junction is identified as the location of the stop-bar. Systematic errors of GPS data are mitigated by the accumulation of trajectory data and data mining of distribution rules. VISSIM simulation results and real-world cases validate the advantage of the proposed recognition method in terms of location errors. A simulation-based sensitivity analysis of the traffic volume, penetration rate, sampling frequency, is conducted to show the robustness of the proposed method. The proposed method can automatically estimate the location of stop-bars of intersections in a network, which makes it possible for large-scale intelligent transportation applications, such as vehicle trajectory planning.

CICTP