ROCAS: A Robust Online Algorithm for Spatial Partitioning in Distributed Smart Camera Systems

Vision based surveillance systems are used for example for the observation of large areas to detect intruders and protect safety critical equipment. These large areas may be maritime ports or aprons of airports and can usually not be guarded solely by human staff due to the costs arising. Most of today´s surveillance systems used for the observation of these areas rely on a centralised system architecture with a central entity being the system´s single point of failure. In this paper, a self-optimising system architecture for networked SCs is proposed. This implies SCs being able to decide autonomously how to arrange their fields of view to achieve an optimal spatial surveillance coverage. Additionally, a distributed algorithm allowing for online spatial partitioning and detection of failing nodes is described and evaluated. The evaluation shows that near optimal spatial partitioning can be achieved without any node having knowledge of all cameras in the system.