Multi-View Vision systems are collaborative distributed applications devoted to image processing.Applications based on Multi-View Vision systems can recognize shapes, track moving targets, etc. to activate alarms, or to propogate information through the network. These systems require Quality of Service from the network and real-time support at the Operating System level.Wireless Sensor Networks (WSNs) were born for monitoring applications providing best effort services in non-critical environments. Nonetheless the recent evolution of hardware platforms, and the existence of real-time kernels as well as network stacks supporting real-time traffic let us envisage the deployment of WSNs to new domains like that of Multi-View Vision.Hereby we discuss the feasibility of an object detection system based on vision and deployed through a WSN. For such a purpose, we implemented in the Real-Time Network Simulator (RTNS) a working model for image detection and in-network processing.Referring to a simple star-shaped network scenario, we analyze the system performances from a real-time perspective.

New1 trends in Wireless Sensor Networks envisage deployments for distributed applications requiring real-time support at the kernel level and Quality of Service at the network level.In this domain, at the design stage, particular attention must be devoted to individual data packets as those entities carrying unique (not redundant) information. The performances of the deployed system (hereby felt as a black box) must be tracked against the reliability and timeliness offered in message delivery.A Visual Tracking case study is discussed throughout this paper with the support of a simulation package modelling real-time scheduling policies at the device node kernels and bandwidth allocation techniques for network reliable communications as standardized in the IEEE 802.15.4 suite of protocols.A set of results is carried out estimating the performances of the Visual Tracking system in two contexts (those of a monitored junction in an airport taxiway and in a parking area) very different for criticality and average volume of network traffic.