A Vision Zero-Oriented Diagnostic Framework for Complex Junctions Using UAV-Based Trajectory Analysis

This study presents a replicable diagnostic framework for analysing latent safety vulnerability at complex junctions by integrating UAV-based observation, trajectory extraction, movement-level operational performance modelling, and regulatory benchmarking. Grounded in Vision Zero/Safe System principles, the approach is demonstrated at Junction 50 of the A-66 motorway in Mieres (Spain), a constrained urban interchange where motorway access/egress overlaps with local cross-town movements. Two one-hour UAV datasets were collected during peak periods and processed with GoodVision to extract trajectories, turning-movement counts, origin–destination patterns and recurrent irregular manoeuvres. These outputs were combined with SIDRA INTERSECTION indicators (e.g., LOS, delay and capacity utilisation) and assessed against the Spanish geometric design standard to identify design–operation misalignment, including a targeted 3D sight-distance check at the stop-controlled exit. The results show systematic behavioural adaptations at critical decision points, including informal side-by-side queuing at nominally single-lane exits and queue bypassing via adjacent auxiliary areas, co-occurring with capacity-strained movements (LOS E–F) and normative inconsistencies in lane type/length and channelisation. The framework translates high-resolution behavioural evidence into intervention-relevant outputs (critical movements, concentration zones and explicit design–behaviour mismatches) without relying on crash frequency as the primary signal, supporting proactive prioritisation in constrained legacy layouts.