地球信息科学学报

地球信息科学学报 ›› 2016, Vol. 18 ›› Issue (3): 406-415.doi: 10.3724/SP.J.1047.2016.00406

• 地球信息科学理论与方法 • 上一篇    下一篇

基于房屋轮廓与纹理的三维建筑模型分层次聚类研究

潘文斌1,2,3(), 刘坡1,4, 周洁萍2, 龚建华2   

  1. 1. 城市空间信息工程北京市重点实验室,北京 100038
    2. 中国科学院遥感与数字地球研究所 遥感科学国家重点实验室,北京 100101
    3. 中国科学院大学,北京 100049
    4. 中国测绘科学研究院,北京100830
  • 收稿日期:2015-05-08 修回日期:2015-06-24 出版日期:2016-03-10 发布日期:2016-03-10
  • 作者简介:

    作者简介:潘文斌(1990-),男,硕士生,研究方向为虚拟地理环境.E-mail:panwb@radi.ac.cn

  • 基金资助:
    国家自然科学基金面上项目(41371387);城市空间信息工程北京市重点实验室开放研究基金项目(2014210)

3D Building Model Hierarchical Generalization Based on Building Contour and Texture

PAN Wenbin1,2,3,*(), LIU Po1,4, ZHOU Jieping2, GONG Jianhua2   

  1. 1. Beijing Key Laboratory of Urban Spatial Information Engineering, Bejing 100038, China
    2. State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing and Digital Earth (CAS), Beijing 100101, China
    3. University of Chinese Academy of Sciences, Bejing 100049, China
    4. Chinese Academy of Surveying & Mapping, Beijing 100039, China;
  • Received:2015-05-08 Revised:2015-06-24 Online:2016-03-10 Published:2016-03-10
  • Contact: PAN Wenbin

RichHTML

PDF (PC)

赞

被引次数

1

摘要：

目前三维建筑模型已广泛应用于城市规划,导航和虚拟地理环境等领域.不同细节的模型是LOD( Level of detail )技术的基础,由于三维模型的生产成本高昂,模型自动化简逐渐引起了学者的关注.三维模型化简包括单模型化简和多模型综合2方面,目前单个模型的化简研究比较多,而模型群组综合的研究仍然处于起步阶段.本文主要研究模型群组的聚类综合,提出一种基于房屋轮廓与纹理的分层次聚类算法:首先,基于房屋的底面轮廓构建约束Delaunay三角网,以道路为基准对三角网进行划分,通过可视分析构建初始的邻接图,使建筑群组分类符合城市形态学;其次,将房屋纹理引入三维模型群聚类的过程,使用SOM( Self-organizing Map )智能分类算法对纹理进行分析,然后分割邻接图;最后,以最邻近距离对邻接图构造最小生成树,并进行线性检测,将离散的建筑合并到已聚类的群组中,最终完成模型的合并.本文利用纹理辅助轮廓特征,实现三维建筑模型的聚类,符合人类的视觉习惯,实验结果证明了本文方法的有效性.

关键词: 三维模型, 综合, 房屋轮廓, 纹理, SOM

Abstract:

As a major factor of smart city, 3D building model has been widely applied in many fields such as city planning, navigation and virtual geographic environments. Level of detail (LOD) technology is often used to visualize these models in complex urban environment. In order to reduce the high production cost, 3D building model generalization is gradually attracting attention. 3D building model generalization includes single model simplification, which has been studied widely and deeply, and multiple models synthesis, which is still at its early stages. In this paper, a self-organizing map (SOM) based classification algorithm is used to analyze the texture, and then the texture is combined with geometry feature to generalize 3D building models. In addition, a cognition-based hierarchical clustering algorithm is proposed. Firstly, a constrained Delaunay triangulation (CDT) is constructed from the 2D footprints of building models, and a connected network is established by visibility analysis based on the CDT, which is divided based on the road network to make it conform to the urban morphology theory. An initial proximity graph is constructed from the CDT. Secondly, the proximity graph is segmented according to the texture feature. Finally, a minimum spanning tree (MST) is created from the proximity graphic based on the minimum distance, and after linear detection and discrete model conflation, the model groups are conflated and visualized. The texture feature is used to synthesize 3D model that is adaptive to people's cognitive habits. The experimental results confirm the effectiveness of this method.

Key words: 3D building models, generalization, building contour, texture, SOM