三维场景模型并行处理策略及简化方法研究

发布时间：2018-04-18 04:01

本文选题：3D城市模型 + 模型简化　；参考：《湖南科技大学》2017年硕士论文

【摘要】：由于大数据、互联网、云计算等技术逐步地运用到GIS中,数字城市时代已经逐渐地转换为智慧型城市时代.与此同时,人们对数据信息分析功能需求不再仅仅满足于某些信息的直观展示功能,而是对知识的可视化程度的提高以及对决策方案的准确性和实时获取的效率等方面更加严格。这在为三维GIS技术的发展提供了新的发展机遇的同时,也给其提出了新的挑战。随着数据量以TB、GB级的迅速增加,计算复杂程度也随时提升,传统的数据库管理软件和传统数据处理技术在可接受的时间内收集、存储、检索、分析、挖掘和可视化能力已经难以满足其要求。在对三维数据进行绘制、查询以及显示等操作时,为了保证稳定而又流畅的互交体验就必须克服计算量大、耗时长等计算缺点。基于以上背景,本文对三维城市模型数据的组织方法和多细节层次模型进行了深入的分析和延展,深化了三维模型的空间应用分析功能。本文的主要工作及创新点如下:(1)提出了一种异构架构的并行处理方法,并考虑到GPU存储器端的存储容量,讨论了异构架构并行方法的基本原理和实际操作构架、并对异构架构中存在的数据的负载衡量和数据传输方面进行了讨论和分析。(2)对经典的模型简化方法进行改进,针对地物与地形集成模型建立一种可控输出模型质量和精度的动态连续LOD简化方法。(3)基于语义识别和视觉认知理论对城市模型进行全局性聚类概括。全局性聚类概括采用道路网的分割方式,充分利用了城市形态学的特征,保证结构不变性。引入拓扑关系、高程阈值、面积阈值、方位阈值等与人类空间认知的视觉特性相关因素,提高了聚类精度。
[Abstract]:As big data, Internet, cloud computing and other technologies are gradually used in GIS, the digital city era has been gradually transformed into the intelligent city era.At the same time, the requirement of data information analysis function is no longer satisfied with the visual display function of some information.It is more strict to the visualization of knowledge, the accuracy of decision-making scheme and the efficiency of real-time acquisition.This not only provides a new opportunity for the development of 3D GIS technology, but also presents a new challenge to it.With the rapid increase in the amount of data in the TBB GB level, the complexity of the calculation also increases at any time. Traditional database management software and traditional data processing technology collect, store, retrieve and analyze the data in acceptable time.Mining and visualization ability has been difficult to meet its requirements.In order to ensure a stable and smooth intersecting experience, the shortcomings of computation such as large computation amount and long time consuming must be overcome when rendering, querying and displaying 3D data.Based on the above background, the method of organizing 3D urban model data and multi-level of detail model are deeply analyzed and extended in this paper, and the spatial application analysis function of 3D model is deepened.The main work and innovation of this paper are as follows: (1) A parallel processing method of heterogeneous architecture is proposed. Considering the storage capacity of GPU memory, the basic principle and practical operation framework of parallel method in heterogeneous architecture are discussed.At the same time, the paper discusses and analyzes the load measurement and data transmission of data in heterogeneous architecture.Based on semantic recognition and visual cognition theory, a dynamic continuous LOD simplification method for quality and precision of controllable output model is established.The method of road network segmentation is adopted in global clustering, which makes full use of the characteristics of urban morphology and ensures the invariance of structure.The related factors such as topological relation, elevation threshold, area threshold, azimuth threshold and visual characteristics of human spatial cognition are introduced to improve the clustering accuracy.
【学位授予单位】：湖南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;P208

【参考文献】