An intelligent system architecture for multi-camera human tracking at airports

In this paper we describe the architecture of an intelligent surveillance system tested at two reference airports. This architecture is developed to support the human operator and enables a multi-camera tracking of suspicious people in case of an alert. The described architecture is based on a network of non-overlapping cameras, each one connected to a self-developed recording tool which provides acquired images to different image processing modules. An efficient preprocessing makes it possible to analyze the data in realtime. The system is able to detect, track and recognize people, but also enables the prediction of where a person will walk to by analyzing possible walking paths.