Airport level-of-service assessment using video analytics: a data-driven framework for airport queue monitoring

Airport terminals require continuous Level of Service assessment, yet commercial computer vision systems provide only operational metrics (counts, dwell times, flows), without systematic connection to service quality frameworks. This gap prevents data-driven decision making for queue management and resource allocation. This research develops a queue Level of Service (qLoS) metric integrating four passenger experience dimensions—queueing time, path distance, queue density, and baggage dynamics—weighted through Analytic Hierarchy Process technique on a respondent panel of 25 effective respondents across two collection rounds. The computer vision pipeline operates on standard overhead Closed-Circuit Television cameras using a custom head detection model (9267 images, ∼463,350 annotations), multi-object tracking via custom RobustTracker, and pole-calibrated height addition projection. A time-weighted density methodology integrates crowding exposure over individual queue residence duration. Hybrid-protocol validation on a 47-passenger airport check-in observation (n = 43 included subset after a pre-specified geometric calibration-envelope exclusion) demonstrated per-passenger queue-time agreement within ±15% for 43/47 passengers overall. Mean absolute percentage error was 4.20% against ground truth on the included subset. Sensitivity analysis across scoring-scale granularity and AHP weight variation over three real respondent-panel scenarios places the scene-mean qLoS in the Over-Design / Optimum boundary region (Pearson’s r = 0.9592 between the 4-score and 6-score scoring systems). Deployment using existing CCTV infrastructure demonstrates a practical path for continuous, data-driven service quality monitoring in airport information management.