Privacy preserving crowd estimation for safer cities
Salvatore Longo, Bin Cheng · 2015
As sensors are getting cheaper and being widely deployed in cities, the Internet of Things (IoT) is providing us great potential to make cities not only smart but also safer. However, how to utilize the information sensed from physical environments to improve civil safety is still a challenging issue. In this paper we present a methodology to estimate the crowd level for indoor scenarios based on data collected from inexpensive and privacy conscious sensors. Our method can be widely applied into many applications for Safer Cities, ranging from facility monitoring to emergency handling. Different from camera-based approaches, our solution can preserve user privacy, scaling better in terms of costs than existing solutions. Using the data collected from real environments, we examine different supervised learning algorithms and identify that Random Forest is the best model. Our solution has been deployed and tested in a Singapore shopping mall, showing that 95% of the overcrowded situations can be detected.