Investigation of traffic conflicts at signalised intersections in Warsaw

Olszewski, Piotr; Czajewski, Witold; Osińska, Beata; Szagała, Piotr; Włodarek, Paweł · 2019 · OpenAlex-citations

DOI: 10.1051/matecconf/201926205009

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the persistent safety risks faced by vulnerable road users (VRUs), specifically pedestrians and cyclists, at signalized intersections in Warsaw, Poland. Despite general improvements in traffic safety, accident rates for VRUs have not significantly decreased, with police records from 2010–2014 indicating 735 pedestrian and 505 cyclist injuries. Approximately 30–33% of these crashes occurred at signalized intersections. The research aims to validate surrogate safety measures as diagnostic tools by establishing a correlation between traffic accidents and traffic conflicts (near-accidents). This approach overcomes the limitations of traditional accident data analysis, which suffers from low event frequency, reporting biases, and long observation periods required for statistical reliability. The investigation was conducted as part of the European Union’s InDeV project. Researchers selected three typical signalized intersections in Warsaw based on historical crash data involving VRUs and turning vehicles. Video recordings were captured using synchronized high-resolution color and thermal cameras. The study utilized two specialized software tools: RUBA (Road User Behaviour Analysis) for semi-automatic detection of road users via virtual motion detectors, and T-Analyst for trajectory analysis and conflict severity assessment. The methodology involved manual alignment of wireframe models to video frames every four seconds to determine trajectories, allowing for the calculation of surrogate safety indicators, primarily Time-to-Collision (TTC) and Post-Encroachment Time (PET). The paper also details technical challenges, including false positives caused by shadows or noise, and false negatives due to occlusion or slow-moving objects. Preliminary results were derived from manual analysis of 24-hour video recordings at two sites (PL3 and PL4) to establish a "ground truth" dataset. At site PL3, 366 vehicle-pedestrian and 69 vehicle-cyclist encounters were identified; at site PL4, 307 vehicle-pedestrian encounters were recorded. TTC values were calculable for approximately 40% of pedestrian encounters, with all values exceeding 2.0 seconds, indicating no immediate collision risk. However, PET analysis revealed significant conflict potential. Encounters with PET values ≤ 1.0 second, indicative of dangerous situations, comprised 1.6% of pedestrian encounters at PL3, 7.2% of cyclist encounters at PL3, and 5.9% of pedestrian encounters at PL4. Cyclists exhibited lower median PET values than pedestrians, suggesting higher risk, likely because vehicles cross cycle lanes before pedestrian crossings. The study concludes that RUBA and T-Analyst are effective tools for identifying and analyzing traffic conflicts, though they currently require significant expert input and manual processing. The findings demonstrate that surrogate safety indicators, particularly PET, can identify dangerous interactions that do not result in recorded accidents. These statistical distributions of encounter parameters provide a foundation for developing robust safety indicators. The authors anticipate that advancements in artificial intelligence and deep learning will eventually enable fully automated conflict detection, facilitating more efficient and objective traffic safety assessments.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• naturalistic crash near crash