| Abstract |
Smart cities in current and future pandemics are expected to implement features that ensure social distancing in order to prevent the spread of infection. Technologies for sensing street-level crowd density are considered helpful in avoiding crowded situations; however, street-level crowd density is tiff& cult to sense effectively using existing techniques. In this paper, we propose a method for sensing street-level crowd density with good accuracy by observing public Bluetooth low energy (BLE) advertisements from popular contact tracing applications. We conducted an experiment in major shopping districts in Tokyo by deploying our developed sensing devices and demonstrated that our method can estimate the street-level crowd density in 30-min intervals with high accuracy, compared to manually counting the number of pedestrians. Using this method, we have begun to publish the street-level crowd density on our website and a news program on Japanese television. Moreover, through long-term monitoring of the collected street-level crowd density data, we analyzed the factors that affect crowd density and constructed a model to predict crowd density from other factors with a coefficient of determination of 0.9 or higher using support vector regression. |
