⚡ Quick Summary
This study introduces a novel framework for real-time crash risk forecasting at signalized intersections by integrating generalized extreme value (GEV) theory with advanced AI-based video analytics. The results indicate that the developed model can accurately predict crash risks, enhancing safety management in traffic systems.
🔍 Key Details
- 📊 Dataset: 97 hours of video footage analyzed
- ⚙️ Technology: Deep neural network-based computer vision for traffic conflict detection
- 📈 Models used: Non-stationary GEV model, ARIMA, GRU, and LSTM for forecasting
- 🏆 Performance: Mean crash frequency estimates within 95% confidence limits of observed crashes
🔑 Key Takeaways
- 🚦 Real-time applications of crash risk forecasting can significantly improve traffic safety.
- 💡 Integration of AI and traffic conflict techniques offers granular insights into crash risks.
- 📉 Forecasting accuracy extends reliably up to 11 future signal cycles.
- 🔍 Non-stationary GEV model effectively estimates opposing-through crash risks.
- 🤖 AI-based video analytics enhances the detection of traffic conflicts.
- 🌍 Potential for broader implementation in intelligent transport systems for proactive safety management.
📚 Background
Traffic safety remains a critical concern, particularly at signalized intersections where the risk of crashes is heightened. Traditional methods of crash risk assessment often rely on historical data, which may not provide the timely insights necessary for effective intervention. Recent advancements in artificial intelligence and traffic sensing technologies present an opportunity to enhance real-time crash risk forecasting, enabling more proactive safety measures.
🗒️ Study
The study conducted by Howlader and Haque focuses on developing a unified framework that integrates generalized extreme value (GEV) theory with various forecasting models to predict crash risks at signalized intersections. By analyzing 97 hours of video footage, the researchers employed a deep neural network-based computer vision technique to extract post encroachment time (PET) traffic conflicts, which are critical for estimating crash risks.
📈 Results
The findings reveal that the mean crash frequency estimates derived from the non-stationary GEV model align closely with the observed crash data, falling within the 95% confidence limits. Furthermore, both autoregressive and recurrent neural network models demonstrated comparable forecasting accuracy, confirming the robustness of the developed framework in predicting crash risks effectively.
🌍 Impact and Implications
The implications of this study are significant for traffic safety management. By leveraging AI and advanced forecasting techniques, the proposed framework can serve as a vital component of intelligent transport systems. This innovation not only enhances the accuracy of crash risk predictions but also facilitates timely interventions, ultimately leading to improved safety outcomes at signalized intersections.
🔮 Conclusion
This research highlights the transformative potential of integrating AI-based video analytics with generalized extreme value theory for real-time crash risk forecasting. As traffic systems evolve, the adoption of such innovative frameworks can pave the way for proactive safety management, significantly reducing the likelihood of crashes and enhancing overall road safety. Continued exploration in this field is essential for further advancements.
💬 Your comments
What are your thoughts on the integration of AI in traffic safety management? We would love to hear your insights! 💬 Share your comments below or connect with us on social media:
Opposing-through crash risk forecasting using artificial intelligence-based video analytics for real-time application: integrating generalized extreme value theory and time series forecasting models.
Abstract
Recent advancements in artificial intelligence (AI) and traffic sensing technologies provide significant opportunities for real-time crash risk forecasting. While forecasting based on historical crash data yields macroscopic insights into future crash risks, such information is often insufficient for real-time applications. In contrast, traffic conflict techniques (TCTs) leveraged by extreme value theory (EVT) and AI-based video analytics have enabled crash risk estimation to a granular level, presenting a promising potential for real-time applications. This study develops a unified framework of integrating generalized extreme value (GEV) theory with parametric and non-parametric forecasting models to predict opposing-through crash risks at signalized intersections. A deep neural network-based computer vision technique was employed to extract post encroachment time (PET) traffic conflicts from 97 h of video footage. Crash risks were estimated using a non-stationary GEV model, incorporating traffic conflict counts, speed variations, and signal timing characteristics. These risk estimates were then forecasted using autoregressive integrated moving average (ARIMA), gated recurrent unit (GRU), and long short-term memory (LSTM) models to analyze short-term crash trends. Results show that the mean crash frequency estimates fell within the 95 % confidence limits of observed crashes and confirm the adequacy of the developed EVT model in estimating opposing-through crashes. The autoregressive and recurrent neural network models exhibit similar forecasting accuracy for crash risk forecasting, with reliable predictions extending up to 11 future signal cycles. The proposed real-time crash risk forecasting framework can be a crucial component of an intelligent transport system, leading to proactive safety management for signalized intersections.
Author: [‘Howlader MM’, ‘Haque MM’]
Journal: Accid Anal Prev
Citation: Howlader MM and Haque MM. Opposing-through crash risk forecasting using artificial intelligence-based video analytics for real-time application: integrating generalized extreme value theory and time series forecasting models. Opposing-through crash risk forecasting using artificial intelligence-based video analytics for real-time application: integrating generalized extreme value theory and time series forecasting models. 2025; 218:108073. doi: 10.1016/j.aap.2025.108073