The Adaptive Collection and Analysis of Distributed Multimedia Sensor Data

Traditional sensor networks communicate all data to a central repository for collection and analysis. The data volumes are low so this works: data are moved. Multimedia sensor networks for environmental monitoring present a different scenario. Volumes of data are large and sensors may be located in areas of limited or poor connectivity. Typical cloud solutions do not work. In this paper we present techniques we have used to collect and analyse data from distributed sensors. These include moving some or all data, moving data summaries and moving computation. Where networked bandwidth is low we employ sneaker net to move data from sensors to servers and from servers to volunteers for analysis, in much the same way as advocated by Gray and Szalay.