集群环境下大范围三维复杂场景的并行绘制机理及算法实现

Drawing depending on single computer with the help of graphics accelerator card and static algorithm has been unable to meet the needs for the growing geographic spatial information and the needs for the aided spatial decision-making support. By means of　data granularity model theory, load balancing theory, image fusion theory of large scale complex three-dimensional scene, efficient real-time visualization of large scope, complex scenes in the parallel rendering theory and method problem. Aiming at the large scope, complex three-dimensional scene drawing in speed, capacity, picture realistic problems and defects were studied, based on PC clusters of large complex scene of the parallel rendering method, the research contents include: ( 1) PC cluster based parallel computing architecture; ( 2) suitable for parallel mapping of the 3D model quick access to spatial database structure and organization, ( 3) PC clusters environment of parallel computing and rendering, including dynamic task allocation equilibrium, real-time frame synchronization, multi channel and large screen image fusion and correction. Research is expected to for a large range, multi-scale complex scene real-time visualization to open up a new way, and then promote the three-dimensional geographic information system and the popularization application of efficient visual decision..Large range, multi-scale three-dimensional scene real-time parallel rendering is restricting the 3D city model ( 3DCM) value-added application . 3DCM as the main content of digital city, it has a variety of levels of detail 3D geometric expression, and provide realistic textures and texture feature and object attribute information, its important, depth application is the aided spatial decision-making support. Although the research of 3DGIS visualization technology has made some achievements, but increasing the velocity of mass data roaming, scene of the real and the sense of beauty, landscape database build are awaited for further studied. Due to the large extent, multi-scale 3DCM construction directly led to massive data problems, and has become the bottleneck of 3DCM dynamic visualization. Based on existing graphics accelerator single drawing, static algorithm can not meet the growing demand for its 3DCM data visualization. Based on the PC cluster parallel rendering is the solution of large-scale complex scene real-time visualization of important means, the core of the method is three: ( 1) PC cluster based parallel computing architecture; ( 2) the 3DCM spatial database structure and data organization; ( 3) the parallel computing algorithm model in the frame of PC cluster architecture.

目前三维场景的单机绘制及其采用的静态算法已无法满足日益增长的三维地理空间信息对其辅助空间决策支持的表达需求。借助粒度模型理论、负载平衡理论、图像融合理论等研究大范围、多尺度三维复杂场景的高效实时可视化问题，突出解决大范围复杂三维场景并行绘制中的理论与方法难题。针对目前大范围、三维复杂场景绘制在速度、容量、画面真实感等方面存在的问题和缺陷进行研究，提出集群环境下大范围三维复杂场景的并行绘制机理、模式、算法，并在小型集群上进行验证性实现与应用，具体研究内容包括：（1）基于PC集群的并行绘制架构与集群通讯机制；（2）大范围复杂3DCM的空间存储与数据组织模型，（3）集群环境下并行绘制算法,包括任务动态分配平衡、实时帧同步、多通道屏幕图像融合与矫正，（4）小型集群原型系统。研究成果有望为大范围、多尺度三维复杂场景的实时可视化开辟一条新的途径，进而促进三维地理信息系统的高效可视化决策与大众化应用。

目前三维场景的单机绘制及其采用的静态算法已无法满足日益增长的三维地理空间信息对其辅助空间决策支持的表达需求。借助粒度模型理论、负载平衡理论、图像融合理论等研究大范围、多尺度三维复杂场景的高效实时可视化问题，突出解决大范围复杂三维场景并行绘制中的理论与方法难题。针对目前大范围、三维复杂场景绘制在速度、容量、画面真实感等方面存在的问题和缺陷进行研究，提出集群环境下大范围三维复杂场景的并行绘制机理、模式、算法，并在小型集群上进行验证性实现与应用，具体研究内容包括：（1）基于PC集群的并行绘制架构与集群通讯机制；（2）大范围复杂3DCM的空间存储与数据组织模型，（3）集群环境下并行绘制算法,包括任务动态分配平衡、实时帧同步、多通道屏幕图像融合与矫正，（4）小型集群原型系统。针对上述问题进行研究，提出了保持视觉特征的LOD模型简化算法；提出了基于Mongo DB 的遥感影像分布式存储策略；实现了基于三维标量场的光线投射并行体绘制算法；等。研究成果包括：发表论文25篇，其中SCI1篇，EI3篇，CSCD—C16篇，其他论文5篇；出版专著 3部；硕士论文6部；取得知识产权15项，其中发明专利4项，实用新型专利11项；获得省部级奖励3项。l研究成果有望为大范围、多尺度三维复杂场景的实时可视化开辟一条新的途径，进而促进三维地理信息的高效可视化及其大众化决策应用。