Application-independent in-network storage optimization for distributed Smart Camera systems

Today´s Smart Camera networks are able to execute a wide variety of distributed algorithms. The usage of high-level algorithms range from image processing, i.e. video analysis and feature extraction, to distributed observation and control protocols that coordinate the collaboration of cameras. These algorithms output georeferenced data representing various abstractions of the observed scene as well as the cameras´ states. This data is again used by other algorithms operating on it or visualizing it to the security personnel. We present a novel approach for optimizing the storage of georeferenced sensor data in Smart Camera Networks. Our approach makes use of an access-centric paradigm to adaptively optimize the storage location of data depending on the associated access patterns. Our storage framework offers a generalized interface to be easily usable by any higher-level algorithm. The algorithm´s adaptive behaviour leads to a huge reduction of network usage and latency.