顾及结构特征的建筑物图形迭代凸分解方法  

An iterative approach for building convex decomposition considering their shape characteristics

作　　者：魏智威 童莹 丁愫 乔海浪 WEI Zhiwei;TONG Ying;DING Su;QIAO Hailang(Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100094,China;Key Laboratory of Network Information System Technology,Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100094,China;School of Resources and Environment Science,Wuhan University,Wuhan 430079,China;College of Environmental and Resource Science,Zhejiang A&F University,Hangzhou 311300,China)

机构地区：[1]中国科学院空天信息创新研究院,北京100094 [2]中国科学院空天信息创新研究院网络信息体系技术院重点实验室,北京100094 [3]武汉大学资源与环境科学学院,武汉430079 [4]浙江农林大学环境与资源学院,杭州311300

出　　处：《测绘科学》2022年第2期184-191,共8页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41871378)。

摘　　要：针对建筑物图形,该文提出了一种结合其形态表达特点的迭代凸分解方法。该方法利用Delaunay三角网提取建筑物图形凹部层次化结构和骨架线定义图形节点凹度,并依据建筑物图形直角化和简洁表达的特点选取分割线迭代消除当前图形中凹度最大节点实现图形凸分解。实验结果表明,该文方法能有效实现带洞(或岛)建筑物图形的凸分解,结果符合视觉认知。同时,基于图形分解结果讨论了字母型和直线型建筑物群模式的识别,证明了结合图形分解能更有效挖掘建筑物群空间分布特征。Representing shapes in terms of meaningful parts is a fundamental problem in shape analysis and recognition.We proposed an iterative convex decomposition approach for buildings considering their shape characteristics.In this approach,node concavity is defined based on built implicit hierarchy of concave parts and skeleton of the building by using Delaunay triangulation.The best cut is then selected to eliminate the node with the largest concavity iteratively to achieve building convex decomposition.Orthogonal features of buildings are considered in the cut selection.The experimental results show that the approach proposed can effectively achieve convex decomposition of buildings with or without holes,and the decomposition results are in line with human visual perception.Letter-like and linear building pattern recognition were also discussed based on decomposition results,which show spatial distribution of building groups can be explored effectively with shape decomposition.

关 键 词：凹度 凸分解 形状认知 模式识别 

分 类 号：P208[天文地球—地图制图学与地理信息工程]