Estimating Crowd Density in an RF-Based Dynamic Environment

Abstract

Crowd density estimating is a crucial service in many applications (e.g., smart guide, crowd control, etc.), which is often conducted using pattern recognition technologies based on video surveillance. However, these kinds of methods are high cost, and cannot work well in low-light environments. Radio frequency based technologies are adopted more and more in indoor application, since radio signal strength (RSS) can be easily obtained by various wireless devices without additional cost. In this paper, we introduce a low cost crowd density estimating method using wireless sensor networks. The proposed approach is a device-free crowd counting approach without objects carrying any assistive device. It is hard to count objects based on RSS measurement, since different number of mobile people at different positions often generates different RSS due to the multipath phenomenon. This paper utilizes the space-time relativity of crowd distribution to reduce the estimation errors. The proposed approach is an iterative process, which contains three phases: the training phase, the monitoring phase, and the calibrating phase. Our experiments are implemented based on TelosB sensor platform. We also do some large-scale simulations to verify the feasibility and the effectiveness of our crowd density estimating approach.