DTC-FM: Expert-Annotated UAV Traffic Conflict and Surrogate Safety Dataset
Year: 2026
Type: dataset
Abstract: This dataset (FME-V) contains 2,508 expert-annotated traffic interaction and surrogate conflict events extracted from high-resolution UAV video trajectories across 11 intersections (collected under the TÜBİTAK 123M640 / KARDS project). It is designed to train, validate, and benchmark classification models (such as the Dual-Threshold Controlled Filter Model - DTC-FM) aimed at mitigating false alarms in Post-Encroachment Time (PET) and trajectory-based surrogate safety analysis. The dataset includes: * Kinematic & Geometric Features: Core and derived trajectory-based metrics (e.g., , , , , velocities , acceleration , conflict angle , and total object count ). · Interaction Class / Sequence: Road user interaction types provided in native categorical (character/string) format representing vehicle and vulnerable road user (VRU) encounter sequences. * Ground Truth Target (): Expert consensus binary labels (1: Surrogate Conflict, 0: Normal Interaction / Undisturbed Pass) established through deliberated expert consensus.
Source: Aperta Turkey Open Archive
Institution: Kırıkkale University
Language: English
Topic: Traffic and Road Safety
Subfield: Safety, Risk, Reliability and Quality
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7203472178