Continuous Monitoring of a Signalized Intersection Using Unmanned Aerial Vehicles

Espadaler-Clapés, Jasso; Fonod, Robert; Barmpounakis, Emmanouil; Geroliminis, Nikolas · 2023 · Crossref

DOI: 10.1109/itsc57777.2023.10422673

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of conducting continuous, full-day traffic monitoring at signalized intersections using unmanned aerial vehicles (UAVs). While drones offer high-resolution trajectory data, their limited battery life typically restricts data collection to short durations, such as peak hours. The authors demonstrate that commercial, low-cost drones can effectively monitor traffic throughout an entire day by utilizing short, sequential flight intervals rather than continuous hovering. The study focuses on a busy intersection in Manchester, UK, to validate this approach and extract comprehensive traffic metrics. The experimental design involved four DJI Mini 3 drones flying at 120 meters altitude to record video footage of five critical intersections over four days in late 2022. To overcome battery constraints, pilots executed multiple short flights, each lasting approximately 3 to 4 minutes, covering morning (07:00–11:00) and afternoon (14:00–17:00) peak hours. The methodology employed a computer vision pipeline using YOLOv5 for object detection and the OC-SORT algorithm for multi-object tracking. Trajectories were georeferenced using homography matrices and smoothed with a linear Kalman filter. Traffic analysis included calculating origin-destination (OD) matrices, turning ratios, fundamental diagrams, and shockwave speeds using virtual loop detectors and critical point identification. The results indicate that brief video segments are sufficient to derive accurate, continuous traffic metrics. OD analysis revealed that the north-south axis was the predominant flow for cars and buses, while trucks primarily utilized the east-west axis. Turning ratios remained stable throughout the day, with north-south movements accounting for approximately 80% of traffic. Fundamental diagrams showed significant congestion at the south entry and north entry during the morning peak, which dissipated by 10:00 AM, whereas the south entry remained congested. Shockwave analysis confirmed that queue formation and dissipation occurred faster in the morning than in the afternoon, with starting shockwaves consistently traveling faster than stopping shockwaves, aligning with traffic flow theory. The significance of this work lies in proving that tethered drones or extended battery solutions are not necessary for continuous intersection monitoring. By leveraging short, frequent flights, researchers can obtain detailed, time-resolved traffic data at a lower cost and with greater operational flexibility. This approach enables the extraction of complex metrics like shockwave propagation and fundamental diagrams, providing a robust framework for future multimodal traffic analysis and smart city applications.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• exposure measurement