Application of Wi-Fi Packet Sensors for Real-Time Visitor Estimation During Large-Scale Events

Abstract

Accurately estimating visitor numbers during events with sudden surges in attendance is essential for transportation management, marketing, and disaster prevention. This study examined the relationship between visitor counts derived from Wi-Fi packet sensor data and two benchmarks: manual pedestrian traffic counts and cellular location information from a third-party service. The temporal resolution of Wi-Fi data and event-specific factors introduced analytical challenges. Given the time-series nature of the data and deviations from normality, Poisson regression with the generalized least squares method was employed. Results demonstrated a statistically significant relationship between Wi-Fi sensor data and pedestrian counts, highlighting the potential of Wi-Fi sensors in tracking real-time crowd dynamics. However, estimation accuracy was inconsistent owing to technical limitations such as signal interference, MAC ID randomization, and overlapping signals. Further research is recommended to improve sensor calibration and integrate advanced machine learning models for enhanced accuracy.