Automating the Estimation of Turning Movement Rates at Multilane Roundabouts Using Unmanned Aerial Vehicles and Deep Learning
Bibliographic record
Abstract
Estimating turning movement rates (TMRs) at multilane roundabouts is more complex than that at conventional intersections due to the vehicle maneuvers and the difficulty of tracking each vehicle through the roundabout. Traditional data collection methods often fail to capture these interactions. With the rise of lightweight sensors, UAVs equipped with cameras provide a comprehensive top-view perspective that street-level cameras lack. Despite these advantages, gaps remain in using UAVs for traffic data collection, particularly focusing on roundabout shapes, and a framework for varying traffic conditions is lacking. This research develops an automated framework using UAVs to collect traffic data at multilane roundabouts. This study proposes an automated framework integrating UAV-based video capture with advanced deep-learning models to improve vehicle detection, tracking, and turning movement estimation at multilane roundabouts. The study evaluates state-of-the-art YOLOv8 architectures for detection and compares DeepSORT and ByteTrack for multi-object tracking, enabling accurate trajectory extraction. The research aims to (1) enhance detection and tracking accuracy using advanced deep learning models that were among the most robust and widely adopted at the time of the study, (2) develop a computationally efficient TMR estimation algorithm based on UAV-extracted trajectories, and (3) assess the framework’s performance under different conditions and its generalizability across sites. Results showed that the YOLOv8l model balances precision, recall, and efficiency, making it the preferred model for detection. DeepSORT outperformed ByteTrack in tracking accuracy and efficiency. Both demonstrated high transferability when applied to new datasets, maintaining high mAP50 scores and excellent identity tracking. The TMRs estimation, verified against manual counts, achieved 97% accuracy. This study contributes a scalable and efficient framework for UAV-based traffic analysis by deploying advanced deep-learning models for accurate vehicle detection and trajectory extraction. Specifically, the integration of YOLOv8 and DeepSORT demonstrated strong performance in real-world multilane roundabout scenarios, underscoring their potential for enhancing data-driven traffic operations and decision-making. • Developed a UAV-ML-based framework for estimating turning movement rates at multilane roundabouts. • YOLOv8l vehicle detection model balances accuracy and computational efficiency. • DeepSORT outperformed ByteTrack in terms of vehicle tracking accuracy and efficiency. • The proposed framework achieved 97% accuracy, verified by manual counts. • The proposed framework shows high transferability to different roundabouts.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".