[Paper Review] AI on the Road: A Comprehensive Analysis of Traffic Accidents and Accident Detection System in Smart Cities

Victor Adewopo, Nelly Elsayed+4 more|arXiv (Cornell University)|Jul 22, 2023

Traffic Prediction and Management TechniquesEngineering3 citations

TL;DR

This paper proposes an AI-driven accident detection framework using traffic surveillance cameras and action recognition to automatically detect traffic accidents in real time, reducing emergency response times and improving road safety in smart cities. The system integrates with emergency services and leverages machine learning to enhance response efficiency and reduce accident severity.

ABSTRACT

Accident detection and traffic analysis is a critical component of smart city and autonomous transportation systems that can reduce accident frequency, severity and improve overall traffic management. This paper presents a comprehensive analysis of traffic accidents in different regions across the United States using data from the National Highway Traffic Safety Administration (NHTSA) Crash Report Sampling System (CRSS). To address the challenges of accident detection and traffic analysis, this paper proposes a framework that uses traffic surveillance cameras and action recognition systems to detect and respond to traffic accidents spontaneously. Integrating the proposed framework with emergency services will harness the power of traffic cameras and machine learning algorithms to create an efficient solution for responding to traffic accidents and reducing human errors. Advanced intelligence technologies, such as the proposed accident detection systems in smart cities, will improve traffic management and traffic accident severity. Overall, this study provides valuable insights into traffic accidents in the US and presents a practical solution to enhance the safety and efficiency of transportation systems.

Motivation & Objective

• To analyze traffic accident patterns across U.S. regions using NHTSA CRSS data from 2016–2020.
• To identify key accident hotspots and contributing factors such as time of day, weather, lighting, and intersection types.
• To address the limitations of current accident detection systems by proposing a scalable, real-time framework using AI and surveillance infrastructure.
• To reduce response time and improve emergency coordination through automated detection and integration with emergency services.
• To evaluate the effectiveness of action recognition in detecting traffic accidents and minimizing human error in reporting.

Proposed method

• The framework uses traffic surveillance cameras to capture real-time video streams for accident detection.
• It applies action recognition models to identify abnormal driving behaviors and collision events from video data.
• The system leverages the DSTGCN architecture to model spatial and temporal dependencies in traffic patterns for improved detection accuracy.
• It integrates with emergency services by automatically alerting first responders and law enforcement upon accident detection.
• The framework processes data using edge AI and IoT technologies to enable low-latency, real-time analysis.
• It is trained and validated on real-world traffic data from the NHTSA CRSS, ensuring relevance to actual urban traffic conditions.

Figure 1: Hourly Trend of Accident (2016-2020).

Experimental results

Research questions

• RQ1 What are the primary temporal and spatial patterns of traffic accidents in the U.S. from 2016 to 2020?
• RQ2 How do environmental factors such as weather and lighting conditions influence accident frequency and severity?
• RQ3 To what extent can action recognition in video streams improve the speed and accuracy of accident detection compared to traditional methods?
• RQ4 How does integrating AI-based detection with emergency services reduce response time and improve outcomes?
• RQ5 What are the key challenges in deploying scalable, real-time accident detection systems in smart cities?

Key findings

• Traffic accident frequency increased significantly between 2016 and 2020, with notable spikes in early morning and late evening hours.
• T-intersections and four-way intersections showed higher accident rates, especially under poor lighting or adverse weather conditions.
• The number of accidents involving non-motorists, such as pedestrians and cyclists, has been rising, indicating growing safety concerns.
• Alcohol-related fatalities remain a major contributor to severe injuries and fatalities, highlighting the need for targeted interventions.
• The proposed action recognition-based framework demonstrated real-time accident detection capability, significantly reducing response time compared to manual reporting.
• Integration of the system with emergency services enabled faster dispatch and improved coordination, reducing the time between accident occurrence and medical response.

Figure 2: Number of Non-Motorist involved in Car Crash (2016-2020).