多样用户行为不确定条件下的实时全景可视化数据传输研究

Panoramic visualization has been an important application foreground in the field of Intelligent Transportation, Tele-education and Emergency Management. However, due to the large volume of panoramic visualization data collected by multiple panoramic cameras, the efficiency of the panoramic visualization data processing and transmission has been significantly reduced. Additionally, the complexity of user behavior and the diversity of user devices in rendering, computing, and bandwidth have made the existing methods of the panoramic visual data transmission difficult to apply to the real-time application of panoramic visualization. This could dramatically restrict the feasibility of real world and object reconstruction in a virtual environment. We observe that it is mainly due to the redundancy of the data locating and transmission. To some extent，not all the panoramic visualization data are essential. In this project, we mainly focus on the study of user behavior and the panoramic visualization data real-time transmission technology. To facilitate it, a novel user behavior character analysis and behavior prediction theoretical model is constructed regarding the user information needs psychological and the behavioral rules. With this model, the studies on the adaptive strategy of user logical topology organization in the virtual environment, the data catching method based on user behavior prediction and multi-resolution model of the real-time transmission method are designed, which is expected to solve the problem of the panoramic visualization data real-time transmission fundamentally. The proposed theoretical models and techniques in this project can be applied to the real-time video /data applications that have stringent requirements on the transmission latency, which is very important and valuable.

全景可视化在智能交通、远程教育和应急管理等领域都有着重要的应用前景。由于现场多个相机的存在，大范围的采集易对全景可视化数据的处理和传输都造成很大障碍；同时由于用户行为复杂及用户设备在绘制、计算和网络等方面有很大的差异，现有全景可视化数据的传输方法，存在数据查找和传输冗余，难以应用于大规模复杂场景的全景可视化，无法实现对真实环境和对象的实时重现。本项目以全景可视化数据实时传输技术研究为核心，深入研究全景可视化应用中用户行为的特点，根据用户信息需求心理和行为规律建立基于用户行为特征的分析和预测理论模型；研究虚拟空间内的自适应用户逻辑拓扑组织方法、基于用户行为预测的全景数据预缓存方法和基于感知的多分辨率模型实时传输方法，从而从根本上解决全景可视化中实时性和扩展性低的问题。本项目研究的理论模型与技术能够用于实时类应用的视频和模型数据传输，具有重要的意义和价值。

可视化数据实时传输问题限制了多相机全景可视化技术的发展和广泛应用。本项目围绕多样用户行为不确定条件下的实时全景可视化数据传输进行了深入研究。首先，为了挖掘全景可视化中用户行为规律和用户访问模式，给出了用户访问行为定义和提出了用户访问模式发现方法。针对网络动态性和用户行为的不确定性，构建了多层次网络逻辑拓扑。利用高性能用户的计算、绘制能力和网络通信能力为域内节点提供数据分发服务，降低了数据请求传输时间和提高了系统的可扩展性。提出了基于用户行为预测的全景数据预缓存方法，能够根据用户访问模式的群体性和差异性分配权重，结合用户访问模式和当前访问场景，自动预测和缓存全景数据。针对多路相机视频区域重叠度低情况，提出了基于视频模型的虚实融合方法，其中视频模型是根据图像内容和三维场景关系计算重构出视频模型，实现图像场景的视频模型生成。提出了基于感知的多分辨率模型实时传输方法，实现接收端多分辨率模型实时传输和模型重建，能够动态地适应终端接收端性能。本项目搭建了相应的理论框架，提出和实现了面向多用户的实时全景数据传输方法及相关关键技术相关取得了一定的成果，研究成果可应用于社会安全、远程教育、智能交通和应急管理等领域。