A Middleware of Things for supporting distributed vision applications

What is a smart camera? Which capabilities has it in terms of on-board processing and networking? In this paper we discuss about design and implementation issues of Wireless Multimedia Sensor Networks exploiting the potential of autonomous low power devices in terms of flexibility and reconfigurability. Considering a large scale scatter of camera nodes connected to the Internet via an IPv6-like suite of protocols we can enable these devices with a set of computer vision primitives, composable as required by the system. In this scenario a WMSN does not represent an application-specific entity (like a system specialized on tracking), but a framework providing a set of diversified services derived by the composition of those primitives. We will discuss some results obtained implementing computer vision techniques over microcontrollerbased embedded systems. We suggest a possible architecture for a Middleware of Things able to handle data, events, and code; this design will permit the deployment of flexible distributed algorithms, dependent on the environmental context including directives by operators and the occurrence of specific events.