Building shape encoding model integrating skeleton and contour features

doi:10.11947/j.AGCS.2026.20260175

Abstract

Abstract:

Building shape plays an important role in the cartographic generalization of buildings, and is essential for building feature selection, simplification, shape preservation, and multi-scale representation. Existing methods mostly perform modeling based on local or regional features of buildings. Although they can characterize local geometric variations in building contours, they lack effective constraints on global structure, making it difficult to simultaneously preserve contour details and global morphological characteristics. This limitation hinders the overall representation capability regarding complex building forms. This paper proposes an unsupervised shape representation learning network, skeleton-contour shape representation network model (SCSR-Net), which integrates building skeleton and contour features. By constructing a skeleton point-set autoencoder to learn skeleton features reflecting the global structure of buildings and a graph convolutional autoencoder to learn contour features reflecting local building details, the model achieves joint learning and fusion through a joint encoding and feature aggregation mechanism, which yields shape encodings that capture both skeleton and contour characteristics of building shapes. Experimental results demonstrate that the SCSR-Net effectively integrates global and local information, overcomes the limitations of single-feature encoding, and achieves excellent performance in shape matching and retrieval tasks, thereby validating the effectiveness of the proposed model.

Key words: building shape representation, feature fusion, cartographic generalization, unsupervised learning

CLC Number:

P208

Xiechun Lu, Peiyang Hou, Kai Ma, Zhanlong Chen, Yongyang Xu. Building shape encoding model integrating skeleton and contour features[J]. Acta Geodaetica et Cartographica Sinica, 2026, 55(9): 1679-1694.

Add to citation manager EndNote|Reference Manager|ProCite|BibTeX|RefWorks

URL: http://xb.chinasmp.com/EN/10.11947/j.AGCS.2026.20260175

http://xb.chinasmp.com/EN/Y2026/V55/I9/1679

Figures/Tables 21

Fig. 1

Fig. 2

Tab. 1

Fig. 3

Fig. 4

Fig. 5

Tab. 2

Fig. 6

Tab. 3

Fig. 7

Fig. 8

Fig. 9

Fig. 10

Tab. 4

Tab. 5

Fig. 11

Tab. 6

Tab. 7

Tab. 8

Tab. 9

Tab. 10

References 39

[1]	Batty M. The size, scale, and shape of cities[J]. Science, 2008, 319(5864): 769-771.
[2]	Arefi M. Urban transformation: understanding city design and form[J]. Journal of Urban Design, 2011, 16(2): 304-306.
[3]	Deng Min, Li Zhilin, Chen Xiaoyong. Extended Hausdorff distance for spatial objects in GIS[J]. International Journal of Geographical Information Science, 2007, 21(4): 459-475.
[4]	Ai Tinghua, Cheng Xiaoqiang, Liu Pengcheng, et al. A shape analysis and template matching of building features by the Fourier transform method[J]. Computers, Environment and Urban Systems, 2013, 41: 219-233.
[5]	Ballatore A, Bertolotto M, Wilson D C. An evaluative baseline for geo-semantic relatedness and similarity[J]. GeoInformatica, 2014, 18(4): 747-767.
[6]	Moradi M, Roche S, Mostafavi M A. A novel GIS-based polygon shape similarity measure applied to OSM building footprints[J]. Advances in Cartography and GIScience of the ICA, 2025, 5: 1-9.
[7]	Larson R R. Geographic information retrieval and spatial browsing[C]//Proceedings of the 32nd Clinic on Library Applications of Data Processing. Urbana-Champaign: Graduate School of Library and Information Science, University of Illinois at Urbana-Champaign, 1996: 81-124.
[8]	Veltkamp R C, Hagedoorn M. State of the art in shape matching[M]//Lew M S. Principles of Visual Information Retrieval. London: Springer, 2001: 87-119.
[9]	Zhang X, Ai T H, Stoter J, et al. Data matching of building polygons at multiple map scales improved by contextual information and relaxation[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2014, 92: 147-163.
[10]	Yan Yuhan, Wu Pengda, Yin Yong, et al. Robust multi-source geographic entities matching by maximizing geometric and semantic similarity[J]. Scientific Reports, 2024, 14(1): 31616.
[11]	Yan Haowen. Quantitative relations between spatial similarity degree and map scale change of individual linear objects in multi-scale map spaces[J]. Geocarto International, 2015, 30(4): 472-482.
[12]	Yan Xiongfeng, Ai Tinghua, Zhang Xiang. Template matching and simplification method for building features based on shape cognition[J]. ISPRS International Journal of Geo-Information, 2017, 6(8): 250.
[13]	Sester M. Cartographic generalization[J]. Journal of Spatial Information Science, 2020(21): 5-11.
[14]	武芳, 巩现勇, 杜佳威. 地图制图综合回顾与前望[J]. 测绘学报, 2017, 46(10): 1645-1664. DOI: . doi: 10.11947/j.AGCS.2017.20170287
	Wu Fang, Gong Xianyong, Du Jiawei. Overview of the research progress in automated map generalization[J]. Acta Geodaetica et Cartographica Sinica, 2017, 46(10): 1645-1664. DOI: . doi: 10.11947/j.AGCS.2017.20170287
[15]	钱海忠, 武芳, 朱鲲鹏, 等. 一种基于降维技术的街区综合方法[J]. 测绘学报, 2007, 36(1): 102-107, 118.
	Qian Haizhong, Wu Fang, Zhu Kunpeng, et al. A generalization method of street block based on dimension-reducing technique[J]. Acta Geodaetica et Cartographica Sinica, 2007, 36(1): 102-107, 118.
[16]	Zahn C T, Roskies R Z. Fourier descriptors for plane closed curves[J]. IEEE Transactions on Computers, 1972, C-21(3): 269-281.
[17]	Arkin E M, Chew L P, Huttenlocher D P, et al. An efficiently computable metric for comparing polygonal shapes[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1991, 13(3): 209-216.
[18]	Persoon E, Fu K S. Shape discrimination using Fourier descriptors[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1977, 7(3): 170-179.
[19]	Zhang Dengsheng, Lu Guojun. Review of shape representation and description techniques[J]. Pattern Recognition, 2004, 37(1): 1-19.
[20]	Belongie S, Malik J, Puzicha J. Shape matching and object recognition using shape contexts[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 24(4): 509-522.
[21]	Latecki L J, Lakamper R. Shape similarity measure based on correspondence of visual parts[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(10): 1185-1190.
[22]	晏雄锋, 艾廷华, 杨敏, 等. 地图空间形状认知的自编码器深度学习方法[J]. 测绘学报, 2021, 50(6): 757-765. DOI: . doi: 10.11947/j.AGCS.2021.20210046
	Yan Xiongfeng, Ai Tinghua, Yang Min, et al. Shape cognition in map space using deep auto-encoder learning[J]. Acta Geodaetica et Cartographica Sinica, 2021, 50(6): 757-765. DOI: . doi: 10.11947/j.AGCS.2021.20210046
[23]	Yan Xiongfeng, Ai Tinghua, Yang Min, et al. Graph convolutional autoencoder model for the shape coding and cognition of buildings in maps[J]. International Journal of Geographical Information Science, 2021, 35(3): 490-512.
[24]	Yan Xiongfeng, Yang Min. A comparative study of various deep learning approaches to shape encoding of planar geospatial objects[J]. ISPRS International Journal of Geo-Information, 2022, 11(10): 527.
[25]	Zhou Z Y, Fu C, Weibel R. Move and remove: multi-task learning for building simplification in vector maps with a graph convolutional neural network[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2023, 202: 205-218.
[26]	Yan Xiongfeng, Ai Tinghua, Yang Min, et al. A graph convolutional neural network for classification of building patterns using spatial vector data[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2019, 150: 259-273.
[27]	于洋洋, 贺康杰, 武芳, 等. 面状居民地形状分类的图卷积神经网络方法[J]. 测绘学报, 2022, 51(11): 2390-2402. DOI: . doi: 10.11947/j.AGCS.2022.20210134
	Yu Yangyang, He Kangjie, Wu Fang, et al. Graph convolution neural network method for shape classification of areal settlements[J]. Acta Geodaetica et Cartographica Sinica, 2022, 51(11): 2390-2402. DOI: . doi: 10.11947/j.AGCS.2022.20210134
[28]	张付兵, 孙群, 马京振, 等. 融合全局和局部特征的建筑物形状智能分类方法[J]. 测绘学报, 2024, 53(9): 1842-1852. DOI: . doi: 10.11947/j.AGCS.2024.20240040
	Zhang Fubing, Sun Qun, Ma Jingzhen, et al. An intelligent classification method for building shape based on fusion of global and local features[J]. Acta Geodaetica et Cartographica Sinica, 2024, 53(9): 1842-1852. DOI: . doi: 10.11947/j.AGCS.2024.20240040
[29]	周旭升, 谭永滨, 于忠海, 等. 融合矢量结构特征的复杂建筑物形状识别方法[J]. 测绘学报, 2026, 55(5): 838-849. DOI: . doi: 10.11947/j.AGCS.2026.20250445
	Zhou Xusheng, Tan Yongbin, Yu Zhonghai, et al. A complex building shape recognition method integrating vector structural features[J]. Acta Geodaetica et Cartographica Sinica, 2026, 55(5): 838-849. DOI: . doi: 10.11947/j.AGCS.2026.20250445
[30]	刘鹏程, 黄欣, 马宏然, 等. 建筑物多边形高精度识别的傅里叶形状描述子神经网络方法[J]. 测绘学报, 2022, 51(9): 1969-1976. DOI: . doi: 10.11947/j.AGCS.2022.20210730
	Liu Pengcheng, Huang Xin, Ma Hongran, et al. Fourier descriptor-based neural network method for high-precision shape recognition of building polygon[J]. Acta Geodaetica et Cartographica Sinica, 2022, 51(9): 1969-1976. DOI: . doi: 10.11947/j.AGCS.2022.20210730
[31]	杨雨雪, 王红艳, 张玲玲, 等. 基于细化和最小生成树的多边形主骨架线提取[J]. 上海师范大学学报(自然科学版), 2022, 51(2): 204-209.
	Yang Yuxue, Wang Hongyan, Zhang Lingling, et al. The extraction of main skeleton line of polygon based on thinning and minimum spanning tree[J]. Journal of Shanghai Normal University (Natural Sciences), 2022, 51(2): 204-209.
[32]	罗登瀚, 钱海忠, 何海威, 等. 面状建筑物多层次骨架线提取方法[J]. 测绘科学技术学报, 2019, 36(3): 324-330.
	Luo Denghan, Qian Haizhong, He Haiwei, et al. A method of extracting multi-scale skeleton lines for polygon buildings[J]. Journal of Geomatics Science and Technology, 2019, 36(3): 324-330.
[33]	Li Yiyan, Lu Xiaomin, Yan Haowen, et al. A skeleton-line-based graph convolutional neural network for areal settlements' shape classification[J]. Applied Sciences, 2022, 12(19): 10001.
[34]	禄小敏, 马犇, 闫浩文, 等. 基于骨架线的地图建筑物形状分类方法[J]. 浙江大学学报(工学版), 2024, 58(12): 2479-2488.
	Lu Xiaomin, Ma Ben, Yan Haowen, et al. Shape classifying method of map building based on skeleton line[J]. Journal of Zhejiang University (Engineering Science), 2024, 58(12): 2479-2488.
[35]	Qi C R, Su H, Mo K, et al. PointNet: deep learning on point sets for 3D classification and segmentation[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 77-85.
[36]	Qi C R, Yi L, Su H, et al. Pointnet++: deep hierarchical feature learning on point sets in a metric space[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2017: 5105-5114.
[37]	Lu Xiechun, Li Houpu, Xu Yongyang, et al. Measuring the similarity between shapes of buildings using graph edit distance[J]. International Journal of Digital Earth, 2024, 17(1): 2310749.
[38]	Li Pengbo, Yan Haowen, Lu Xiaomin. A Siamese neural network for learning the similarity metrics of linear features[J]. International Journal of Geographical Information Science, 2023, 37(3): 684-711.
[39]	陈占龙, 刘江文, 徐永洋, 等. 融合深度聚类网络与模板匹配的建筑物化简方法[J]. 测绘科学技术学报, 2025, 41(5): 448-457.
	Chen Zhanlong, Liu Jiangwen, Xu Yongyang, et al. A building simplification method based on deep clustering network and template matching[J]. Journal of Geomatics Science and Technology, 2025, 41(5): 448-457.

特征	指标	计算公式	含义
位置特征	横坐标		其中c_x表示建筑物中心点的横坐标；r表示建筑物最大空间跨度
位置特征	纵坐标		其中c_y表示建筑物中心点的纵坐标
结构特征	节点的度deg_i	-	骨架线顶点的度
轮廓关联特征	最近距离特征		其中ƏB表示轮廓点集
轮廓关联特征	次近距离特征		其中q′表示最近轮廓点

类型	E形	F形	H形	I形	L形	O形	T形	U形	Y形	Z形
模板
数据

形状匹配方法	FT	ST	DCG
转向函数（TF）	0.417	0.526	0.867
傅里叶描述符（FD）	0.726	0.894	0.965
图编辑距离（GED）	0.865	0.977	0.989
图卷积自编码器（GCAE）	0.868	0.934	0.986
LineStringNet模型	0.986	0.997	0.998
深度聚类形状自编码器（DCSAE）	0.923	0.982	0.995
SCSR-Net	0.968	0.995	0.998

融合方法	FT	ST	DCG	匹配准确率/（%）
门控融合（未映射）	0.835	0.966	0.985	91.12
注意力机制（未映射）	0.816	0.920	0.980	91.98
加权拼接（未映射）	0.891	0.982	0.989	94.34
门控融合（映射）	0.871	0.969	0.986	93.22
注意力机制（映射）	0.928	0.993	0.993	94.66
SCSR-Net（映射）	0.968	0.995	0.998	99.18

权重设计	α₁	α₂	α₃	FT	ST	DCG	匹配准确率/（%）
SCSR-Net	动态调整	动态调整	动态调整	0.968	0.995	0.998	99.18
固定权重组1	0.8	0.10	0.10	0.759	0.961	0.953	74.72
固定权重组2	0.7	0.15	0.15	0.738	0.892	0.950	71.78
固定权重组3	0.6	0.20	0.20	0.661	0.838	0.939	70.10
固定权重组4	0.4	0.30	0.30	0.582	0.740	0.911	58.12
固定权重组5	0.3	0.35	0.35	0.536	0.734	0.899	54.60