Device-free Crowd Density Estimation with Off-the-shelf WiFi Traffic

Crowd density estimation plays a crucial role in avoiding potential hazards in public social activities like holiday celebrations. Traditional crowd density estimation approaches leverage cameras to probe the characteristics of the crowd. Despite their high accuracy, they cannot work in bad lighting conditions and could cause visual privacy leakage, To address these problems, in this paper, we propose a WiFi-based crowd density estimation system named WiCount. It is deployed on a side of the entrance of the target public place and continuously collects the communication packets emitted by the access point (AP). By feeding the features extracted from the packets into machine learning classifiers, WiCount is able to determine the number of people passing through the entrance. To deal with the issue of lacking communication traffic for crowd sensing, we design an incentive strategy to drive the AP to transmit compensation packets without impacting the normal communication function too much. Real-world experiments show that WiCount has over 99% and 95% accuracy in detecting the passing through of the people and determining the number of these people.