Drones Spot Traffic Crashes 12 Minutes Faster

Road crashes cause over a million deaths globally each year, with delays in emergency response significantly increasing the risk of fatalities. Current detection methods, like CCTV and dashcams, struggle with low visibility and privacy issues, often missing incidents until it's too late. The DARTS system, developed by researchers, uses drones and AI to detect traffic incidents in real time, offering a faster, more reliable solution that could save lives and reduce congestion.

DARTS achieved 99% accuracy in detecting incidents, such as rear-end collisions, by analyzing drone-captured thermal video streams. It distinguishes between normal traffic, recurrent congestion, and incident-induced disruptions, providing immediate alerts to transportation management centers. In a field test on Interstate 75 in Florida, the system identified a crash 12 minutes earlier than traditional methods, allowing for quicker medical and traffic management responses.

The system integrates drones equipped with thermal cameras, which operate effectively in low-light conditions and avoid capturing sensitive details like license plates. A lightweight deep learning model, called the Traffic Condition Detection Network (TCD-Net), processes vehicle trajectories extracted from video to classify traffic states. This model uses convolutional neural networks enhanced with attention mechanisms to focus on key movement patterns, ensuring high precision without heavy computational demands.

Results from the study show that DARTS not only detects incidents but also tracks their impact, such as congestion length and propagation speed. For example, in the Florida test, it monitored a congestion span that grew from 0.265 miles to 0.5032 miles, propagating at 101.02 feet per minute. The system's web-based interface allows operators to verify incidents visually and access real-time data, streamlining decision-making for rerouting and emergency dispatch.

This innovation matters because faster incident detection can cut emergency response times, potentially reducing severe injuries and fatalities. It also helps mitigate secondary crashes and traffic delays, which contribute to economic losses and environmental pollution. DARTS's flexibility makes it suitable for remote or resource-limited areas, where traditional infrastructure is sparse, supporting global safety goals like the UN Sustainable Development Goals.

Limitations include the need for drones to remain within visual line of sight under current regulations, though the system is designed for beyond-visual-line-of-sight operations with approvals. The study's dataset, while comprehensive, was collected in specific Florida locations, and broader applicability requires testing in diverse environments. Future work could explore multi-drone fleets for wider coverage, but scalability depends on regulatory advancements and infrastructure support.