EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

YoFlow Method for Scenario-Based Automatic Accident Detection

Aditya Haryanto, Ondřej Vaculín

IEEE Open Journal of Intelligent Transportation Systems · 2026 · 7 · 61-73

DOI: 10.1109/ojits.2025.3639557 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Open Journal of Intelligent Transportation Systems, 2026. Verfasst von Aditya Haryanto und 1 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Open Access Publication Fund of Technische Hochschule Ingolstadt, European Union’s Horizon Europe Research and Innovation Funding Programme.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Recent advances in sensor and computing technologies have enabled road side units (RSUs) to not only monitor traffic flow but also process data in real time to improve road safety. However, leveraging RSUs for proactive accident detection remains a challenging and underexplored task, partly due to the lack of diverse accident data. To address this, this study proposes two key contributions: (i) a scenario-based synthetic data generation framework, and (ii) YoFlow, a novel system for vehicle-to-vehicle accident detection from a simulated RSU camera perspective. The proposed framework leverages the PEGASUS method for scenario generation strategy and BeamNG.tech for generating synthetic traffic videos. This approach led to the development of the SB-SIF dataset, which includes five representative intersection crash scenarios derived from German accident data. The SB-SIF dataset contains 914 crash videos, 123 near-miss events, and 924 normal traffic instances and is publicly available at: https://doi.org/10.5281/zenodo.15267252. The proposed YoFlow system identifies accidents by analyzing temporal variations in vehicle speed vectors, using YOLO for vehicle classification and CUDA-accelerated dense optical flow to capture abrupt motion changes. The extracted features are processed and classified using an XGBoost model, achieving 94% recall and 90% precision in accident detection.

Förderung

  • Open Access Publication Fund of Technische Hochschule Ingolstadt
  • European Union’s Horizon Europe Research and Innovation Funding Programme

Zitieren

APA

Haryanto, A., & Vaculín, O. (2026). YoFlow Method for Scenario-Based Automatic Accident Detection. IEEE Open Journal of Intelligent Transportation Systems, 7, 61-73. https://doi.org/10.1109/ojits.2025.3639557

Vancouver

Haryanto A, Vaculín O. YoFlow Method for Scenario-Based Automatic Accident Detection. IEEE Open Journal of Intelligent Transportation Systems. 2026;7:61-73. doi:10.1109/ojits.2025.3639557

BibTeX
@article{Haryanto2026,
  author = {Haryanto, Aditya and Vaculín, Ondřej},
  title = {YoFlow Method for Scenario-Based Automatic Accident Detection},
  journal = {IEEE Open Journal of Intelligent Transportation Systems},
  year = {2026},
  volume = {7},
  pages = {61--73},
  doi = {10.1109/ojits.2025.3639557}
}
Kostenlos starten →