监控视频的三维空间覆盖模型研究

As one kind of the most important sensors in Internet of Things and Earth Observation Sensor Webs, camera can record the spatial and temporal visual information of real world. Nowadays camera deployment and intelligent video analysis have been essential parts of national and industry development planning. However, the current surveillance system cannot produce the best possible results for most monitoring tasks due to lack of overall planning beforehand. A lot of effort has been conducted and many achievements on models and optimizing of surveillance system have been obtained. Although vision is an inherently three-dimensional phenomenon, many coverage models in various applications are two-dimensional with assuming a roughly planar environment. Some works are derived from three-dimensional characteristics of the cameras and scene, but the final representations are two-dimensional. Consequently the state of the art still leaves much room for improvement. The proposal aims to model and optimize of surveillance system in three-dimensional environment to satisfy the actual application requirements. This proposal mainly focuses on following fundamental issues about surveillance system, such as 3D perception model, coverage estimation model and optimal configuration. On one hand this study can enrich the theories and methods for modeling and optimizing of surveillance system. On the other hand, this study can meet the actual demand, because all the surveillance cameras are localized and modeled in 3D geographic environment. Moreover this study can offer feasible theories and methods for detection of blind regions, estimation of coverage ability and optimal configuration of surveillance system. This study can also be used for reference for other directional and omnidirectional sensor networks.

监控视频作为物联网、对地观测传感器网中一类重要的传感器，其部署与内容分析已经纳入到国家和行业发展规划。然而目前建设的监控视频缺乏整体规划，多数设备并未达到最优配置。同时目前研究多将视频传感器感知模型抽象为二维感知平面，且将监控区域抽象为平坦平面，与视频传感器真实感知能力以及实际地理空间不一致，直接导致在此基础上的覆盖分析方法无法直接应用于大规模实用系统。本研究旨在解决三维地理环境下实现监控视频网络优化配置，重点研究视频传感器三维感知模型、监控视频网络覆盖效能模型、优化配置等基础问题。一方面丰富监控视频网络三维感知与覆盖控制的理论与方法；另一方面，将多视频传感器纳入统一的地理空间下进行优化配置，深化监控视频的实用性。同时，本研究可为监控视频网络覆盖盲区检测与效能评估、新建视频传感器优化配置等提供理论与方法支持，并对其他有向、无向传感器的覆盖分析具有指导和借鉴意义。

视频传感器作为物联网、对地观测传感器网中一类重要传感器，可以获取到最为直观、最为丰富的可视可听媒体信息。在重要基础设施安全、社会治安反恐防暴、高危行业安全探测、国土防空等领域有着重要的作用。无论是新建视频监控网络的部署优化，还是已建视频监控网络的调度优化，都急需视频监控系统覆盖能力分析及其优化的理论与方法。本项目旨在通过研究视频传感器三维感知模型、视频传感器覆盖计算、摄像机网络部署优化和调度优化等基础问题，一方面丰富监控视频网络三维感知与覆盖控制的理论与方法；另一方面，将多视频传感器纳入统一的地理空间下进行优化配置，深化监控视频的实用性。.本项目的主要成果如下:在视频传感器感知模型方面，根据CCD/CMOS成像特点和透视投影模型，构建了视频传感器的感知模型；提出了基于单幅监控图像的相机迭代标定方法；研发了“视频传感器时空标签采集系统”。在视频传感器覆盖方面，提出了顾及障碍物的二值格网扫描算法用以快速精确计算视频传感器的覆盖范围，并研发了“视频传感器覆盖范围分析系统”；在视频传感器部署优化方面，研究了视频传感器的感知强度与感知特性，分析了视频传感器姿态变化对感知范围和感知强度的影响；以覆盖率和感知强度为基本参数，基于A*搜索算法，提出了视频传感器的优化部署方法，形成了“视频传感器覆盖仿真系统”；在视频传感器网络调度方面，以覆盖路网的长度、路段数以及参与调度的摄像机数量为优化目标，提出了多目标摄像机路网调度优化方法。在以上工作的基础上，本项目还开展了地理视频方面的拓展研究，实现了监控视频中运动目标与GIS融合方法、地理视频的时空检索方法。.本项目研究成果可为视频传感器感知模型、监控视频网络覆盖盲区检测与效能评估、新建视频传感器优化配置、已有网络调度优化等提供理论与方法支持，并对其他有向、无向传感器的覆盖分析具有指导和借鉴意义。