地球信息科学学报 ›› 2021, Vol. 23 ›› Issue (4): 551-563.doi: 10.12082/dqxxkx.2021.200288

• 地球信息科学理论与方法 •    下一篇

“分形-视距”(FVD)多分辨率纹理生成及组织方法

包馨1,2(), 周国清1,2,*(), 何宏昌1,2   

  1. 1.桂林理工大学广西空间信息与测绘重点实验室,桂林 541006
    2.桂林理工大学测绘地理信息学院,桂林 541006
  • 收稿日期:2020-03-21 修回日期:2020-08-14 出版日期:2021-04-25 发布日期:2021-06-25
  • 通讯作者: 周国清
  • 作者简介:包 馨(1995— ),女,山东临沂人,硕士生,主要从事纹理影像处理及三维模型可视化研究。E-mail: baoxin@glut.edu.cn
  • 基金资助:
    国家自然科学基金项目(41431179);国家自然科学基金项目(41961065);广西创新驱动发展专项(AA18118038);广西创新驱动发展专项(AA18242048)

Study on Fractal Viewing Distance (FVD) Multi-Resolution Texture Data Generation and Organization

BAO Xin1,2(), ZHOU Guoqing1,2,*(), HE Hongchang1,2   

  1. 1. Guangxi Key Laboratory of Spatial Information and Geomatics, Guilin University of Technology, Guilin 541004, China
    2. College of Geomatics and Geoinformation, Guilin University of Technology, Guilin 541004, China
  • Received:2020-03-21 Revised:2020-08-14 Online:2021-04-25 Published:2021-06-25
  • Contact: ZHOU Guoqing
  • Supported by:
    National Natural Science Foundation of China(41431179);National Natural Science Foundation of China(41961065);Guangxi Innovative Development Grand Program(AA18118038);Guangxi Innovative Development Grand Program(AA18242048)

摘要:

纹理数据是三维建筑模型中的重要组成部分,通常会消耗大量的内存,导致模型较难实现流畅稳定的动态可视化。针对这个问题,本文提出了一种面向三维模型可视化的“分形-视距”(FVD)多分辨率纹理生成及纹理组织方法。该方法首先利用分形对纹理进行压缩并将其存入纹理数据库,然后按照视距及目标立面的可见性情况创建和组织对应纹理:当视距符合纹理调用条件且目标立面可见时,根据不同视距,执行多分辨率纹理创建,按照各类三维模型数据间的属性关系建立索引,存入关系数据库;当视距不符合纹理调用条件但该立面又可见时,利用纹理分割算法创建颜色纹理树,将节点合并表示立面纹理。最后根据变化的视角和距离调用相应纹理,实现三维模型的动态流畅可视化。本文选择山东省济南市历下区作为研究区域,使用的纹理数据共5290幅。实验结果表明,与3ds Max、SketchUp Pro相比,本文提出的方法减少内存占用分别约为36.5%和49.1%,帧速率提高约7%和10.9%,极大地提高了大场景三维模型可视化的渲染效率和速率。本文提出的方法在三维建筑模型渲染效率和视觉质量之间取得了很好的平衡,这一问题的改善将为建立智慧城市提供快速有效的科学支撑。

关键词: 三维建筑模型, 多分辨率纹理, 纹理组织, 分形压缩, 纹理分割算法, 视距, 颜色纹理树, 动态可视化

Abstract:

Texture is one of the important information in the three-Dimensional (3D) city photorealistic mapping, which usually occupies large computer memory for storage. As a result, the effective modeling and organization of texture data become an important issue in geographic information science community. Hence, this paper proposes the Fractal Viewing Distance (FVD) multi-resolution texture data generation and organization method for 3D city building photorealistic mapping. First, a fractal compression method based on the self-similarity and self-affine properties of texture is developed, which compresses and stores texture and 3D building model data in the database. Second, the corresponding texture is created and organized according to the visual distance and the visibility of the target facade. When the visual distance meets the texture calling condition and the target facade is visible, the texture data is iteratively decompressed, and the number of texture resampling is determined to create multi-resolution texture data. A texture index is established and stored in a database according to the relationship between various types of 3D building model data. When the visual distance does not meet the texture calling condition, but the target facade is visible, the building facade texture is represented using a colored texture tree created in this paper. The method smooths the texture and divides it into different rectangles as nodes based on the color difference, and then a colored texture tree is built to store these nodes. In the process of model visualization, the nodes are merged to represent the texture. Finally, the corresponding texture is selected according to the angle of view and distance, representing the dynamic and smooth visualization of the 3D model. The experimental field consisting of a total of 5290 texture data, located in Lixia District, Jinan City, Shandong Province is selected to validate the method we proposed, which occupies a total size of 6.1G. The results demonstrate that when the number of decompressing iterations reaches around 8 or 10, the peak ratio of signal to noise of each texture tends to be stable. The average memory consumption is approximately 54.7%, and the frame rate for building fa?ade mapping is at about 42 frames per second. Compared with 3ds Max and SketchUp Pro, the proposed method reduces the memory occupation by about 36.5 % and 49.1%, respectively, and increase the frame rate approximately by 7 % and 10.9 %, respectively. The method proposed in this paper has achieved a good balance between rendering efficiency and visual quality of 3D building model. The method proposed in this paper can provide an effective scientific support for the establishment of smart city.

Key words: 3D building model, multi-resolution texture, texture organization, fractal compression, texture segmentation algorithm, viewing distance, color texture tree, dynamic visualization