基于多媒体传感器网络的建筑灾难现场三维全景实时重建关键问题研究

项目名称： 基于多媒体传感器网络的建筑灾难现场三维全景实时重建关键问题研究

项目编号： No.61273078

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 吴成东

作者单位： 东北大学

项目金额： 82万元

中文摘要： 针对灾难现场信息采集存在的技术难题，提出将新兴的无线多媒体传感器网络与图像处理技术有机结合的灾难现场三维全景实时重建解决方案。为满足灾难现场信息获取对网络容错性、准确性及实时性要求，对灾难现场传感信息处理与通信关键性理论与技术问题进行研究，提出面向动态拓扑结构的多媒体传感器网络通信容错保障机制和基于事件驱动的网络多节点协同高效感知方法，解决网络通讯与信息融合的瓶颈性难题。将三维全景实时重建分解为静态背景重建与人体目标建模，提出基于多目视觉信息与建筑空间模型的静态场景重建方法与基于人体形态与运动模式鲁棒特征的目标实时检测识别算法，研究基于虚拟现实技术的静态背景模型与动态人体模型的融合方法，实现灾难现场动态场景三维全景实时重建。研究开发仿真与物理实验平台，验证方案的先进性和有效性。项目的目标是实现灾难现场信息采集理论与技术新突破，为无线多媒体传感器网络和灾难救援领域提供新的研究思路和关键技术。

中文关键词： 无线多媒体传感器网络；灾难信息感知；目标定位与识别；效能优化；三维优化调度

英文摘要： To solve the technical problems in the information acquisition for disaster sites, we propose a new method for three-dimensional (3-D) panoramas reconstruction of disaster sites based on wireless multimedia sensor networks (WMSN) and image processing. By considering the demands of fault-tolerance, robustness and real time processing, we focus on the key issues of information processing and communication, and propose the WMSN communication fault-tolerant guarantee mechanism for dynamic topological structure and event-driven cooperated effective sensing method of multiple networked nodes, and try to break the bottleneck of the development of network communication and information fusion techniques. The 3-D panoramas real-time reconstruction is divided into two parts: static background reconstruction based on the fusion of multiple-views visual information and building structure model, and real-time object detection and recognition based on the robust features of human body morphology and motion pattern. The integration of the two parts is also studied for realizing the real time reconstruction of the dynamic scenes of building disaster. Besides, we construct a experimental platform for the visual simulation and physical test, which can be used to verify the proposed methods and the research conclusions. The aim of

英文关键词： Wireless Multimedia Sensor networks(WMSN)；disaster information sensing；objective localization and recognition；utility optimization；3D optimization and scheduling