Crowd detection and occupancy estimation through indirect environmental measurements

In this paper, the real-time estimation of indoor occupancy has been evaluated starting from the processing of environmental parameters. The proposed system is based on a low cost and non-invasive wireless sensor network infrastructure for the distributed acquisition of simple quantities like temperature and humidity. The proposed algorithm has been structured in three successive phases of detection, filtering, and classification in order to face with the high variability and instability of such quantities. The problem of occupancy estimation has been recast as an inverse problem solved by means of a customized learning-by-example strategy (i.e., the classification phase). A measurement campaign in a museum area has been carried out and preliminary experimental results have been obtained in order to assess the potentialities and limitations of the proposed strategy.