Enhancing Highway Safety: Accident Detection on the A9 Test Stretch Using Roadside Sensors

2025-02-01

AI Summary

To address the challenge of real-time crash detection on expressways, this paper proposes an edge-side collaborative hybrid framework integrating rule-based and learning-based methods. It fuses high-frame-rate multimodal data from four roadside cameras and a 10-Hz LiDAR, coupled with a lightweight rule engine grounded in trajectory anomaly detection and kinematic constraints, alongside a spatiotemporal graph neural network. We introduce and open-source the first real-world highway crash sequence dataset—comprising over 290,000 2D bounding boxes, 93,000+ 3D bounding boxes, and cross-frame trajectory IDs—annotated in OpenLABEL format. Evaluated on France’s A9 motorway, our system achieves a 92.7% crash detection rate, an average response latency of 1.8 seconds, and a false alarm rate below 0.3 per hour. This work establishes the first reproducible, benchmarkable edge-side collaborative solution for crash detection in open-road environments.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Learning on the Edge & Model Compression
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Road traffic injuries are the leading cause of death for people aged 5-29, resulting in about 1.19 million deaths each year. To reduce these fatalities, it is essential to address human errors like speeding, drunk driving, and distractions. Additionally, faster accident detection and quicker medical response can help save lives. We propose an accident detection framework that combines a rule-based approach with a learning-based one. We introduce a dataset of real-world highway accidents featuring high-speed crash sequences. It includes 294,924 labeled 2D boxes, 93,012 labeled 3D boxes, and track IDs across 48,144 frames captured at 10 Hz using four roadside cameras and LiDAR sensors. The dataset covers ten object classes and is released in the OpenLABEL format. Our experiments and analysis demonstrate the reliability of our method.

Problem

• Highway Safety
• Traffic Accident Detection
• Rapid Response

Innovation

• Hybrid Rule-and-Learning-Based System
• Large-scale Real-world Accident Dataset
• OpenLABEL Format Sharing