An online shadowed clustering algorithm applied to risk visualization in Territorial Security

The identification and processing of the risk sources that prevail in a sensor-monitored area is crucial to guarantee the uninterrupted and efficient operation of the surveillance system. In particular, a time-varying schematic depiction of the risk associated with each object will allow the human expert to draw meaningful conclusions about the system dynamics. In this paper, we introduce an online clustering algorithm for risk visualization in a Territorial Security environment. The clustering machinery leans upon shadowed sets due to their robustness and interpretability. The proposed algorithm is able to process data arriving in real time as it only memorizes a small subset of them. It is strong to noisy and abnormal samples and represents each cluster as a shadowed set. Experiments conducted in a simulated Critical Infrastructure Protection scenario confirm the feasibility and robustness of the proposed technique.