Networks of embedded smart cameras are an emerging technology for a broad range of pervasive computing applications including smart rooms, intelligent infrastructures and security. Smart cameras combine video sensing, processing and communication within a single embedded device and provide sufficient onboard infrastructure to perform high-level video analysis tasks. This paper deals with middleware services required for the efficient deployment and operation of distributed smart cameras. We

more »

... s here on services for autonomous and dynamic reconfiguration. Dynamic reconfiguration refers to the exchange of software tasks as well as the alteration of QoS-levels these tasks provide during runtime. Dynamic reconfiguration provides several advantages over statically configured networks including (i) modification of functionality after deployment and during runtime, (ii) adaptation of the network to changes in its internal and external state, and (iii) better exploitation of the available resources. We have developed the services for dynamic reconfiguration using policies. Policies help to specify rules for the reconfiguration process. By evaluation the policy the new task-level configuration of the network is computed. The reconfiguration is implemented using mobile agents in order to achieve a flexible and scalable middleware service. Our policy-based middleware is demonstrated by a surveillance application. Keywords: pervasive computing; smart cameras; policy-based middleware; embedded systems. Recently much effort has been put into the development of distributed vision systems with smart cameras as key components. Smart cameras are equipped with high-performance onboard computing and communication devices. They combine video sensing, processing and communication within a single embedded device [1] . Networks of distributed