Skip to Main Content
The increasing demand in homeland security speeds up the development of innovative and non-invasive systems to localize and track moving objects in complex environments. In this paper the real-time localization of transceiver-free targets is addressed by means of learning by example methodology that exploits the received signal strength indicator available at the nodes of a wireless sensor network as input data. This approach uses neither dedicated sensors nor active devices put on the target to localize both idle and moving objects. The definition of a customized classifier during an offline training procedure enables the real-time generation of a probability map of presence by processing the output of the support vector machine. Some selected experimental results validate the effectiveness of the proposed methodology applied in real scenarios.