基于多源数据的可视模型与环境构建及其动态仿真

The multi-source data-enabled visual modeling, virtual environment construction, and their tight and seamless integration with dynamic scene simulation, have become the most notable bottleneck that has prevented the widespread popularities and penetration of new virtual reality techniques into general public's everyday life such as cultural consumption and digital entertainment in this information era. It is also evident that this subject is one of the key scientific frontier topics that could enable the development of next-generation virtual reality technologies. Based on the organic fusion of multi-source heterogeneous data and physical model, the most fundamental goals of this NSFC project are to systematically address a series of challenging theoretical issues, discover new knowledge through novel theoretic insights, and offer a suite of high-performance engineering solutions that are urgently needed in the intelligent modeling and interactive simulation of dynamic natural scenes (the time-varying phenomena include complex fluids, intelligent crowds, etc.). Our anticipated research outcomes include information fusion and joint knowledge discovery of multi-source heterogeneous natural scene data, the intelligent construction of 3D models and their interactions in virtual environments, the dynamic evolution of joint-data-physics-coupled hybrid models, multi-source data and knowledge driven virtual environment construction, hybrid model based dynamic simulation, the independent evolution of virtual environment, etc. To achieve the aforementioned ambitious goals, our research activities span from the novel theory and technology innovation to the algorithm and software development. Key criteria to be utilized to judge this project's overall success include: the entire multi-source data scale of the scene (at least 10-million geometric primitives, tens of thousands objects, etc.), time performance for the dynamic scene simulation (at least 30 FPS), the supporting number of simultaneous-interaction players (at least 20 persons), etc. The significant research and system outcomes of this research project will serve as a critical impetus for the new development in entertainment industry, which will also greatly contribute to our nation's key innovation of fundamental research and core technologies in virtual reality fields.

基于多源数据的可视模型与环境构建及其动态仿真，是制约文化信息消费普及的主要问题，也是引领下一代虚拟现实技术发展的基础科学问题。项目以包含复杂流体、智能人群等动态对象的自然场景智能建模和交互仿真为切入点，有效融合多源异构数据与物理模型，研究面向自然环境理解的多源数据联合知识发现、多源数据驱动的混合建模、动态模型自主演化与自然交互三个关键科学问题，创新多源数据语义知识驱动的环境构建、混合建模仿真、虚拟环境自主演化与交互三方面理论和技术，建立一套复杂动态现象智能建模理论和技术体系。预期成果可实现由千万级仿真图元、万个实体构成的复杂自然环境的高真实感实时动态仿真（30帧/秒以上），并支持不少于20人的同时动态交互，可促进娱乐产业快速发展、提高我国虚拟现实领域的源头创新能力。

基于多源数据的可视模型与环境构建及其动态仿真，是制约文化信息消费普及的主要问题，也是引领下一代虚拟现实技术发展的基础科学问题。项目研究严格按照计划进行，完成了预定的研究任务，以包含复杂流体、智能人群等动态对象的自然场景智能建模和交互仿真为切入点，有效融合多源异构数据与物理模型，研究面向自然环境理解的多源数据联合知识发现、多源数据驱动的混合建模、动态模型自主演化与自然交互三个关键科学问题，创新多源数据语义知识驱动的环境构建、混合建模仿真、虚拟环境自主演化与交互三方面理论和技术，建立一套复杂动态现象智能建模理论和技术体系。项目实现了由千万级仿真图元、万个实体构成的复杂自然环境的高真实感实时动态仿真（30帧/秒以上），并支持不少于20人的同时动态交互，可促进娱乐产业快速发展、提高我国虚拟现实领域的源头创新能力。相关研究成果在IEEE TVCG、TIP、ICCV、AAAI等顶级期刊和会议发表、录用高水平论文144篇（SCI论文79篇，Q1区SCI 26篇），申请（含授权）国家发明专利41项，软件著作权2项，在国际国内学术会议报告45人次，培养在读博士和硕士研究生63人。