面状居民地聚类方法的对比分析  

Comparative Analysis of Clustering Methods of Planar Settlements

作　　者：孟妮娜[1] 冯建华 贾钰涵 MENG Ni’na;FENG Jianhua;JIA Yuhan(School of Geoengineering and Mapping,Chang’an University,Xi’an 710054,China)

机构地区：[1]长安大学地质工程与测绘学院,陕西西安710054

出　　处：《测绘地理信息》2023年第3期116-120,共5页Journal of Geomatics

基　　金：国家自然科学基金(41501498);陕西省自然科学研究计划(2021JM-155)。

摘　　要：探索建筑物的空间分布模式信息是建筑物地图综合过程中不可或缺的一部分,以建筑物距离为基础,结合建筑物的大小、形状、方向3种特征因子,将多个聚类算法应用于多边形建筑物的聚类分析,并通过不同的城市街区实地数据集对多个聚类算法进行比较分析。结果表明:k-means算法效率最高,但只能识别近似于球形的群组,对呈线性分布的建筑物模式识别效果较差;具有噪声的基于密度的空间聚类(density-based spatial clustering of applications with noise,DBSCAN)算法可以发现任意形状的集群,其对参数的选择过于敏感,难以从复杂的建筑物群中识别出连贯的群组;具有噪声的基于分层的密度聚类(hierarchical DBSCAN,HDBSCAN)算法可以发现任意形状和密度的群组,但对边界区域的建筑物群识别效果较差;最小生成树(minimum spanning tree,MST)算法能够识别出不同类型的建筑物群模式,但难以确定复杂建筑物群的合理划分阈值。Exploring the spatial distribution patterns of buildings is an integral part of map in map generalization of buildings.We apply multiple clustering algorithms in the clus‐ter analysis of polygon buildings by combining three character‐istic factors of buildings including size,shape and orientation based on the distance of buildings.And we compare and ana‐lyze several clustering algorithms through the field data sets of different city blocks.The results show that k‐means algorithm has the highest efficiency,but it can only identify the groups that are close to the sphere,and the recognition effect of build‐ing patterns with linear distribution is poor.Density‐based spatial clustering of applications with noise(DBSCAN)algo‐rithm can find clusters of arbitrary shape,but it is difficult to identify coherent groups from complex buildings because of its over-sensitivity to parameter selection.Hierarchical DB‐SCAN(HDBSCAN)can find groups of arbitrary shape and density,but the recognition effect of building groups in bound‐ary areas is poor.Minimum spanning tree(MST)algorithm can identify different types of building cluster patterns,but it is still difficult to determine a reasonable threshold for com‐plex building clusters.

关 键 词：地图综合 空间聚类 建筑物群 分布模式 

分 类 号：P283[天文地球—地图制图学与地理信息工程]