Real-time multi-view vision systems using WSNs

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.

多视角视觉系统是致力于图像处理的协作式分布式应用。基于多视角视觉系统的应用能够识别形状、跟踪移动目标等，以触发警报，或通过网络传播信息。这些系统需要网络提供服务质量，并在操作系统层面提供实时支持。 无线传感器网络（WSNs）是为在非关键环境中提供尽力服务的监测应用而诞生的。然而，硬件平台的近期发展，以及实时内核和支持实时流量的网络协议栈的存在，让我们设想将无线传感器网络部署到像多视角视觉这样的新领域。 在此，我们讨论一个基于视觉且通过无线传感器网络部署的物体检测系统的可行性。为此，我们在实时网络模拟器（RTNS）中实现了一个用于图像检测和网内处理的工作模型。 参照一个简单的星型网络场景，我们从实时角度分析系统性能。