Road network grid pattern analysis using a pre-trained model fusing spatial and topological information

doi:10.11947/j.AGCS.2026.20250350

Abstract

Abstract:

Road network pattern recognition is a classical problem in the fields of cartography and geospatial information science, and is widely applied in tasks such as map generalization, spatial cognition, and vector data updating. The extraction and analysis of road network pattern features constitute a crucial step in reflecting urban landscape characteristics and enabling intelligent map operations. However, existing methods face two major challenges: first, a heavy reliance on large amounts of high-quality labeled data; second, an incomplete representation of the spatial-topological information of road networks. As typical unstructured graph data, road networks cannot be directly processed by pre-training methods designed for regular grid data. To address these issues, this paper proposes a self-supervised pre-training model that integrates Euclidean spatial proximity and topological adjacency information. Method ologically, on one hand, a graph convolutional kernel network is employed to encode local path structures, enhancing topological awareness; on the other hand, a spatial attention bias mechanism is designed to incorporate Euclidean distance information into graph node representations, achieving an effective fusion of dual spatial information and thus better capturing the characteristics of road networks. In terms of the learning paradigm, a pre-training strategy is adopted, learning universal representations using only unlabeled road network data, thereby reducing dependence on labeled data. Experimental results on public urban road network datasets demonstrate that the proposed model achieves significant advantages in the road network pattern classification task: an accuracy of 88.03%, surpassing the optimal baseline method (CRHD) by 2.29 percentage points; precision and recall reach 88.89% and 87.47%, respectively, with improvements of 3.19 and 1.57 percentage points; and anF₁ score of 0.878 0, an increase of 0.020 9. This paper validates the effectiveness of the fusion architecture and the pre-training strategy. The proposed model can learn more generalizable intrinsic representations of road networks, providing foundational model support for downstream tasks.

Key words: road network pattern, spatial cognition, map analysis, graph pre-training, graph structure learning

CLC Number:

P208

Wenhao YU, Ziyi ZENG, Yifan ZHANG, Haizhong QIAN. Road network grid pattern analysis using a pre-trained model fusing spatial and topological information[J]. Acta Geodaetica et Cartographica Sinica, 2026, 55(3): 415-424.

Figures/Tables 9

Fig. 1

Fig. 2

Fig. 3

Fig. 4

Fig. 5

Tab. 1

Tab. 2

Fig. 6

Tab. 3

References 25

[1]	胡胜, 王振华, 邢汉发, 等. 中国城市路网形态与交通碳排放关系及影响机制研究[J]. 地球信息科学学报, 2025, 27(7): 1687-1703.
	HU Sheng, WANG Zhenhua, XING Hanfa, et al. Relationship and impact mechanisms between urban road-network form and transport carbon emissions in China[J]. Journal of Geo-Information Science, 2025, 27(7): 1687-1703.
[2]	王安东, 武芳, 巩现勇, 等. 一种城市路网多层次复合网格模式识别方法[J]. 测绘学报, 2023, 52(11): 1994-2006. DOI: . doi: 10.11947/j.AGCS.2023.20220528
	WANG Andong, WU Fang, GONG Xianyong, et al. A recognition approach for compound grid pattern of urban road networks[J]. Acta Geodaetica et Cartographica Sinica, 2023, 52(11): 1994-2006. DOI: . doi: 10.11947/j.AGCS.2023.20220528
[3]	吴小芳, 杜清运, 徐智勇. 多层次移位原则的道路与建筑物空间冲突处理[J]. 测绘学报, 2010, 39(6): 649-654.
	WU Xiaofang, DU Qingyun, XU Zhiyong. Disposal of spatial conflict between roads and buildings based on the multilevel displacement principles[J]. Acta Geodaetica et Cartographica Sinica, 2010, 39(6): 649-654.
[4]	苏世亮, 向娟, 杜清运, 等. 叙事地图学的概念框架构建[J]. 测绘学报, 2025, 54(8): 1371-1388. DOI: . doi: 10.11947/j.AGCS.2025.20250107
	SU Shiliang, XIANG Juan, DU Qingyun, et al. Re-conceptualizing narrative cartography[J]. Acta Geodaetica et Cartographica Sinica, 2025, 54(8): 1371-1388. DOI: . doi: 10.11947/j.AGCS.2025.20250107
[5]	BENGUIGUI L, DAOUD M. Is the suburban railway system a fractal?[J]. Geographical Analysis, 1991, 23(4): 362-368.
[6]	杜清运, 况路路, 任福, 等. 自动驾驶高精度地图特征分析及发展展望[J]. 地球信息科学学报, 2024, 26(1): 15-24.
	DU Qingyun, KUANG Lulu, REN Fu, et al. Feature analysis and development prospect of high definition maps for autonomous driving[J]. Journal of Geo-Information Science, 2024, 26(1): 15-24.
[7]	LÜ Zheng, SUN Qun, MA Jingzhen, et al. Road network generalization method constrained by residential areas[J]. ISPRS International Journal of Geo-Information, 2022, 11(3): 159.
[8]	BARTHÉLEMY M. Spatial networks[J]. Physics Reports, 2011, 499(1): 1-101.
[9]	YEH A G O, ZHONG T, YUE Y. Hierarchical polygonization for generating and updating lane-based road network information for navigation from road markings[J]. International Journal of Geographical Information Science, 2015, 29(9): 1509-1533.
[10]	CRUCITTI P, LATORA V, PORTA S. Centrality measures in spatial networks of urban streets[J]. Physical Review E, 2006, 73(3): 036125.
[11]	田晶, 艾廷华, 丁绍军. 基于C4.5算法的道路网网格模式识别[J]. 测绘学报, 2012, 41(1): 121-126.
	TIAN Jing, AI Tinghua, DING Shaojun. Grid pattern recognition in road networks based on C4.5 algorithm[J]. Acta Geodaetica et Cartographica Sinica, 2012, 41(1): 121-126.
[12]	何亚坤, 艾廷华, 杜欣, 等. 网络空间向量剖分法识别城市路网网格模式[J]. 武汉大学学报(信息科学版), 2018, 43(1): 138-144.
	HE Yakun, AI Tinghua, DU Xin, et al. Grid pattern recognition in street network space by vector tessellation method[J]. Geomatics and Information Science of Wuhan University, 2018, 43(1): 138-144.
[13]	王米琪, 艾廷华, 晏雄锋, 等. 图卷积网络模型识别道路正交网格模式[J]. 武汉大学学报(信息科学版), 2020, 45(12): 1960-1969.
	WANG Miqi, AI Tinghua, YAN Xiongfeng, et al. Grid pattern recognition in road networks based on graph convolution network model[J]. Geomatics and Information Science of Wuhan University, 2020, 45(12): 1960-1969.
[14]	LI Qimai, HAN Zhichao, WU Xiaoming. Deeper insights into graph convolutional networks for semi-supervised learning[C]//Proceedings of 2018 AAAI Conference on Artificial Intelligence. New Orleans: AAAI Press, 2018, 32(1): 3538-3545.
[15]	WANG Zehong, LIU Zheyuan, MA Tianyi, et al. Graph foundation models: a comprehensive survey[EB/OL]. [2025-08-30]. https://arxiv.org/abs/2505.15116.
[16]	高晓蓉, 闫浩文, 禄小敏. 多尺度地图空间居民地语义相似度计算方法[J]. 测绘学报, 2022, 51(1): 95-103. DOI: . doi: 10.11947/j.AGCS.2022.20210074
	GAO Xiaorong, YAN Haowen, LU Xiaomin. Semantic similarity measurement for building polygon aggregation in multi-scale map space[J]. Acta Geodaetica et Cartographica Sinica, 2022, 51(1): 95-103. DOI: . doi: 10.11947/j.AGCS.2022.20210074
[17]	YING Chengxuan, CAI Tianle, LUO Shengjie, et al. Do transformers really perform bad for graph representation?[EB/OL]. [2025-08-30]. https://arxiv.org/abs/2106.05234.
[18]	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Proceedings of the 31st Conference on Neural Information Processing Systems. New York: Curran Associates, 2017.
[19]	余华飞, 邱天奇, 周哲, 等. 图Transformer支持下的河网模式识别[J]. 测绘学报, 2024, 53(11): 2075-2085. DOI: . doi: 10.11947/j.AGCS.2024.20230582
	YU Huafei, QIU Tianqi, ZHOU Zhe, et al. Drainage pattern recognition supported by graph Transformer[J]. Acta Geodaetica et Cartographica Sinica, 2024, 53(11): 2075-2085. DOI: . doi: 10.11947/j.AGCS.2024.20230582
[20]	CHEN Wangyang, HUANG Huiming, LIAO Shunyi, et al. Global urban road network patterns: unveiling multiscale planning paradigms of 144 cities with a novel deep learning approach[J]. Landscape and Urban Planning, 2024, 241: 104901.
[21]	KLATZKY R L. Allocentric and egocentric spatial representations: definitions, distinctions, and interconnections[M]//Spatial Cognition. Berlin: Springer, 1998: 1-17.
[22]	佘玉梅, 熊汉. 模糊模式识别方法研究[J]. 云南民族大学学报(自然科学版), 1998, 7(1): 16-22.
	SHE Yumei, XIONG Han. Research on method of fuzzy pattern recognition[J]. Journal of Yunnan Minzu University (Natural Sciences Edition), 1998, 7(1): 16-22.
[23]	STEINLEY D, BRUSCO M J. Initializing K-means batch clustering: a critical evaluation of several techniques[J]. Journal of Classification, 2007, 24(1): 99-121.
[24]	REYNOLDS Douglas. Encyclopedia of biometrics[M/OL]. Boston: Springe, 2009. http://doi.org/10.1007/978-0-387-73003-5_196.
[25]	DABIRI S, LU C T, HEASLIP K, et al. Semi-supervised deep learning approach for transportation mode identification using GPS trajectory data[J]. IEEE Transactions on Knowledge and Data Engineering, 2020, 32(5): 1010-1023.

路网结构	微调数据集	测试数据集	合计
网格型	2702	676	3378
非网格型	4016	1063	5079
合计	6718	1739	—

方法	准确率/（%）	精确率/（%）	召回率/（%）	F₁值
CRHD	85.74	85.70	85.90	0.857 1
SECA	53.08	51.46	51.06	0.483 3
K-means	71.30	71.34	71.47	0.712 7
GMM	49.63	50.97	50.87	0.488 4
SVM	67.28	67.13	66.72	0.667 7
本文模型	88.03	88.89	87.47	0.878 0

方法	是否预训练	拓扑编码	空间编码	准确率/（%）	精确率/（%）	召回率/（%）	F₁值
无预训练方法	否	否	否	66.76	74.59	68.75	0.652 7
预训练方法	是	否	否	80.68	82.63	81.60	0.806 1
拓扑编码方法	是	是	否	81.31	83.73	80.22	0.805 0
空间编码方法	是	否	是	81.77	82.75	81.03	0.813 0
本文模型	是	是	是	88.03	88.89	87.47	0.878 0