基于GPU的全球多分辨率地形数据实时呈现技术研究

发布时间：2018-11-18 19:04

【摘要】：如何高效地组织和管理全球海量多分辨率地形数据,并利用有限的计算机硬件处理能力实现更好的实时渲染效果一直是全球地形可视化技术研究的核心问题。本文从地球离散网格模型入手,结合现代图形硬件的新特性和现有系统存在的缺陷,系统地研究了基于GPU的全球多分辨率地形数据的实时呈现技术。主要研究内容与研究成果包括:1)全球海量多分辨率地形数据的组织与管理方法研究。首先,通过总结现有全球网格系统的特点和缺陷,结合图形硬件的新特性和高效实时渲染技术对地球离散网格模型的要求,选择了HEALPix地球离散网格模型作为本文研究的基础。然后,根据应用的需要对原模型的存储和索引方式进行了改进,将模型可存储的最高分辨率数据由原来的0.012米提高至7.07×10-13米,可采用图形硬件支持的压缩算法,将所需存储空间至少降至原来的1/3。最后,依据改进后的模型设计了全球海量多分辨率地形数据的组织和管理方案。2)基于GPU的全球地形数据实时绘制算法研究。结合改进的HEALPix模型和现有地形实时绘制算法,本文设计了一套基于改进HEALPix模型的全球多分辨率地形数据实时绘制方案。该方案将地形高程数据作为纹理数据存入到显存中,实时绘制时由GPU硬件算法生成地形网格,然后由运行在GPU中的着色器程序完成三维地形的渲染过程。在该方案中,绘制过程几乎完全在GPU中完成,充分地利用了GPU的硬件网格细分功能和GPU高度并行化特性,从而显著提高了场景渲染速度,降低了对内存和CPU资源的消耗。对比现有方案,该方案解决了南北极极点附近的畸变问题,提高了实时渲染的速度和效果,并简化了渲染引擎自身的复杂度。3)原型系统设计与实现。为了验证本文提出的方案,本文设计并实现了全球地形数据管理系统和全球地形实时绘制系统,使用前者建立了全球地形高程数据库和全球地形影像数据库,使用后者对建立的数据库进行实时绘制和模型测试,得到了较为理想的结果。
[Abstract]:How to efficiently organize and manage global massive multi-resolution terrain data and make use of limited computer hardware processing ability to achieve better real-time rendering effect has been the core problem of global terrain visualization technology. Based on the discrete grid model of the earth and the new characteristics of the modern graphics hardware and the shortcomings of the existing systems, this paper systematically studies the real-time rendering technology of global multi-resolution terrain data based on GPU. The main research contents and results are as follows: 1) the organization and management of global massive multi-resolution terrain data. Firstly, by summing up the characteristics and defects of the existing global grid system, combining the new characteristics of graphics hardware and the requirements of efficient real-time rendering technology for the discrete grid model of the earth, we select the HEALPix Earth discrete grid model as the foundation of this paper. Then, the storage and index method of the original model is improved according to the need of application. The maximum resolution data stored in the model can be increased from 0.012 meters to 7.07 脳 10 ~ (-13) meters. The compression algorithm supported by graphics hardware can be used. Reduce the required storage space to at least a third of the original. Finally, according to the improved model, the organization and management scheme of global massive multi-resolution terrain data is designed. 2) the real-time rendering algorithm of global terrain data based on GPU is studied. Combined with the improved HEALPix model and the existing real-time terrain rendering algorithm, this paper designs a real-time rendering scheme of global multi-resolution terrain data based on the improved HEALPix model. In this scheme, topographic elevation data is stored as texture data in the display memory, and the terrain grid is generated by the GPU hardware algorithm in real-time rendering, and then the rendering process of the 3D terrain is completed by the shader program running in GPU. In this scheme, the rendering process is almost complete in GPU, which makes full use of the hardware grid subdivision of GPU and the high parallelization of GPU, thus greatly improving the rendering speed of scene and reducing the consumption of memory and CPU resources. Compared with the existing schemes, the scheme solves the distortion problem near the poles of the Arctic and South Pole, improves the speed and effect of real-time rendering, and simplifies the complexity of the rendering engine itself. 3) the design and implementation of the prototype system. In order to verify the proposed scheme, this paper designs and implements the global terrain data management system and the global terrain real-time rendering system. Using the former, the global terrain elevation database and the global terrain image database are established. The latter is used for real-time rendering and model testing of the established database, and satisfactory results are obtained.
【学位授予单位】：北京工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P209;TP391.41

【相似文献】