面向突发事件现场监测的全向感知无线多媒体传感器网络

Field surveillance of emergent events is a highly-focused hotspot and difficult point at home and abroad. As a newly emerging subject, wireless multimedia sensor network(WMSN) can provide very promising solution. However, WMSN is hard to be deployed and used at the emergent field due to its directional sensing property. In this project, multiple cameras are introduced to build the omni-directional WMSN, which can solve the limitation of ordinary directional sensor node and be used to make the surveillance network for emergent events. In this project, the omni-vision of multiple cameras will be researched. With the sensing distance decided by specific duty and accuracy requirement, the sensing model of omni-WMSN node can be acquired by fusing the scale sensor and vision sensing. Through proper classification of active level for WMSN node, the energy consumption model will be built on the basis of working sequence and status-shift probability. We aill research the rapid deployment method of the WMSN nodes under the given coverage ratio and working life. The moving anchor node is employed to rapidly calibrate the pose and angle of the WMSN node. Target detection algorithm will also be researched under the dynamic background. To achieve the energy balance and accurate tracking, we will design energy efficient strategy for moving target tracking through nodes collaboration. Combining with the event acquisition, the dynamic situation map can be built with the WMSN network topology. The aim of this project is to make a breakthrough of WMSN theory and technique in the field surveillance of emergent events and supply a feasible solution.

突发事件现场监测是国内外高度关注的热点和难点问题，新兴的无线多媒体传感器网络（WMSN）提供了极具前途的解决方法。然而，受限于有向感知特性，WMSN很难在突发事件现场部署和使用。本项目引入多摄像机构建全向WMSN感知节点，克服常规有向节点的缺陷，构建面向突发事件现场监测的WMSN网络。项目研究多摄像机节点全向视图生成，依据任务和精度的要求确定感知距离，融合标量传感器与视觉感知构建WMSN节点模型；研究WMSN节点状态分级，结合工作时序及状态转移概率建立能耗预测模型；研究给定覆盖率与工作寿命条件下的WMSN节点快速部署方法，采用移动信标点实现WMSN节点位姿的快速标定；在动态背景下目标检测算法研究的基础上，研究能量有效的运动目标多节点协同跟踪算法实现能量均衡和准确跟踪，进而结合WMSN网络拓扑构建现场态势图。项目的目标是实现突发事件现场监测领域的WMSN理论与技术突破，形成可行的解决方案。

无线多媒体传感器网络非常适合于突发事件的现场监测，但节点观测范围及部署过程存在较大的缺陷。本项目引入多摄像机构建全向WMSN感知节点，克服常规有向节点的缺陷，构建面向突发事件现场监测的WMSN网络。针对多摄像机WMSN节点以视觉感知为主的特点，建立了面向任务和精度要求的节点概率感知模型。提出了多摄像机相邻图像的配准算法，实现动态条件下感知质量均衡的全向视觉，并构建了标量传感器与全向视觉融合的WMSN节点复合事件感知模型。有别于传统的WSN仅需要考虑生命期和成本的情况，多媒体传感器节点的覆盖需要考虑时间（能耗）、空间、置信度等多重约束，本项目WMSN从事件栅栏覆盖出发，探讨了感兴趣区域目标分辨率的WMSN节点视角关系，建立了多重约束条件下的复合事件栅栏覆盖优化机制。由于突发事件现场监测通常要求在极短的时间内完成节点部署，并迅速形成动态监测网络实现数据采集。本项目提出以旋翼无人机投放和部署全向WMSN节点，并以携带GPS装置的无人机构成移动信标节点，对所部署的WMSN网络节点进行快速标定，由此研究了WMSN节点的事件部署策略，设计了四旋翼无人机信标节点的航线，并构建了有利于WMSN节点快速标定的光电标识。全向感知WMSN节点部署后，通过协同工作的方式检测现场存在的动态目标或事件。本项目的全向感知WMSN节点由4~6路摄像机构成，而且多个节点之间也构成了协同感知，因而呈现为典型的多摄像机协同工作。考虑到广角摄像机存在的畸变因素，项目提出了视野分界线帮助提高全向感知WMSN节点的目标检测效果，并建立了基于时空约束的跨摄像机多运动目标关联算法和运动目标的多摄像机协同跟踪算法。在理论研究的基础上，采用嵌入式多核处理器分别构建了4路、6路全向感知WMSN节点，并结合云服务开发了开放型的实验系统。这些研究成果的取得，可为全向感知多媒体传感器网络的研究提供有益的助力，能够为突发事件现场的实时、准确、全面监测提供提供有效的理论依据和工程指导，具有重要的科学和工程价值。