Fine structure extraction method of overpass based on trajectory and remote sensing image fusion

doi:10.11947/j.AGCS.2026.20260159

Abstract

Abstract:

Interchanges are critical components of urban grade-separated transportation networks, and the extraction of their fine-grained structures is essential for high-precision map construction and navigation route planning. Existing trajectory-based road network construction methods are susceptible to trajectory sparsity and projection overlap between upper-and lower-level roads, often resulting in road discontinuities and erroneous inter-level connections. Remote-sensing-image-based methods, constrained by two-dimensional orthographic projection and occlusion, have difficulty accurately representing the hierarchical topology of interchanges. To address these issues, this paper proposes a fine-grained interchange structure extraction method that integrates crowdsourced trajectories and remote sensing imagery. First, the crowdsourced trajectory data are preprocessed, and an initial road centerline is generated through density raster construction, skeleton extraction, and vectorization. The centerline is then sampled at equal intervals to construct initial geometric nodes. Next, to restore road network discontinuities caused by sparse trajectories, trajectory-derived geometric nodes are used as geometric priors and fused with candidate nodes extracted from remote sensing imagery to form a unified candidate node set. SAM-Road image features are subsequently employed to infer local connectivity and restore missing road connections. Finally, a topology correction method based on geometric continuity is developed to address pseudo-planar intersections in vertically overlapping areas. Pseudo-nodes are identified by evaluating pairs of oppositely directed extension vectors using cosine similarity, and the upper-and lower-level road networks are topologically decoupled through pseudo-node removal and path reconstruction. Experiments conducted using crowdsourced trajectories and remote sensing imagery from Beijing demonstrate that the proposed method effectively extracts fine-grained interchange structures. The geometric accuracy (GEO-F₁ score) and topological correctness (TOPO-F₁ score) reach 0.922 0 and 0.933 1, respectively, and the overall extraction performance surpasses that of the compared single-source methods.

Key words: overpass, crowdsourced trajectory data, remote sensing image, topological correction, road network construction

CLC Number:

P208

Yali Li, Yuezhu Hao, Longgang Xiang, Caili Zhang, Maohua Liu. Fine structure extraction method of overpass based on trajectory and remote sensing image fusion[J]. Acta Geodaetica et Cartographica Sinica, 2026, 55(8): 1482-1495.

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

URL: http://xb.chinasmp.com/EN/10.11947/j.AGCS.2026.20260159

http://xb.chinasmp.com/EN/Y2026/V55/I8/1482

Figures/Tables 12

Fig. 1

Fig. 2

Fig. 3

Fig. 4

Fig. 5

Fig. 6

Fig. 7

Tab. 1

Tab. 2

Tab. 3

Fig. 8

Tab. 4

References 29

[1]	Yang Xue, Yang Mingchun, Cao Yanjia, et al. Three-dimensional structure determination of grade-separated road intersections from crowdsourced trajectories[J]. International Journal of Applied Earth Observation and Geoinformation, 2023, 125: 103598.
[2]	邓媛媛, 向隆刚, 焦凤伟. 基于正逆向追踪与融合的立交桥结构提取[J]. 地球信息科学学报, 2025, 27(10): 2373-2386.
	Deng Yuanyuan, Xiang Longgang, Jiao Fengwei. Extraction of overpass structure based on forward and backward tracking and fusion of crowdsourced trajectory[J]. Journal of Geo-information Science, 2025, 27(10): 2373-2386.
[3]	焦凤伟, 向隆刚, 邓媛媛, 等. 轨迹局部和长序特征结合的立交桥路网构建方法[J]. 测绘学报, 2025, 54(5): 950-962. DOI: . doi: 10.11947/j.AGCS.2025.20240130
	Jiao Fengwei, Xiang Longgang, Deng Yuanyuan, et al. Constructing grade-separated junctions based on combination of local and long-term trajectory feature[J]. Acta Geodaetica et Cartographica Sinica, 2025, 54(5): 950-962. DOI: . doi: 10.11947/j.AGCS.2025.20240130
[4]	张彩丽, 向隆刚, 李雅丽, 等. 基于出租车轨迹的可导航路网构建[J]. 测绘学报, 2021, 50(12): 1650-1662. DOI: . doi: 10.11947/j.AGCS.2021.20200470
	Zhang Caili, Xiang Longgang, Li Yali, et al. Construction of navigable road network based on taxi trajectory[J]. Acta Geodaetica et Cartographica Sinica, 2021, 50(12): 1650-1662. DOI: . doi: 10.11947/j.AGCS.2021.20200470
[5]	Deng Min, Huang Jincai, Zhang Yunfei, et al. Generating urban road intersection models from low-frequency GPS trajectory data[J]. International Journal of Geographical Information Science, 2018, 32(12): 2337-2361.
[6]	Stanojevic R, Abbar S, Thirumuruganathan S, et al. Robust road map inference through network alignment of trajectories[C]//Proceedings of 2018 SIAM International Conference on Data Mining. Philadelphia: Society for Industrial and Applied Mathematics, 2018.
[7]	Tang Jianbo, Deng Min, Huang Jincai, et al. A novel method for road intersection construction from vehicle trajectory data[J]. IEEE Access, 2019, 7: 95065-95074.
[8]	He Songtao, Bastani F, Abbar S, et al. RoadRunner: improving the precision of road network inference from GPS trajectories[C]//Proceedings of the 26th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems. Seattle: ACM Press, 2018: 3-12.
[9]	胡瀚, 向隆刚, 王德浩. 出租车轨迹数据的道路提取[J]. 测绘通报, 2018(7): 53-57.
	Hu Han, Xiang Longgang, Wang Dehao. Road extraction based on taxi trajectory data[J]. Bulletin of Surveying and Mapping, 2018(7): 53-57.
[10]	杨云. 改进U-Net网络的遥感影像道路提取[J]. 测绘与空间地理信息, 2025, 48(8): 110-113.
	Yang Yun. Improved U-Net network remote sensing image road extraction[J]. Geomatics and Spatial Information Technology, 2025, 48(8): 110-113.
[11]	Batra A, Singh S, Pang Guan, et al. Improved road connectivity by joint learning of orientation and segmentation[C]//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE, 2019.
[12]	Zhang Zhengxin, Liu Qingjie, Wang Yunhong. Road extraction by deep residual U-net[J]. IEEE Geoscience and Remote Sensing Letters, 2018, 15(5): 749-753.
[13]	江娜, 李吉元, 杨景玉, 等. 基于拓扑增强的深度学习遥感影像道路提取方法[J]. 兰州交通大学学报, 2023, 42(2): 45-54.
	Jiang Na, Li Jiyuan, Yang Jingyu, et al. Road extraction method of deep learning remote sensing image based on topology enhancement[J]. Journal of Lanzhou Jiaotong University, 2023, 42(2): 45-54.
[14]	Mattyus G, Luo Wenjie, Urtasun R. DeepRoadMapper: extracting road topology from aerial images[C]//Proceedings of 2017 IEEE International Conference on Computer Vision. Venice. IEEE: 2017: 3458-3466.
[15]	He Songtao, Bastani F, Jagwani S, et al. Sat2Graph: road graph extraction through graph-tensor encoding[C]//Proceedings of 2020 European Conference on Computer Vision. Glasgow: Springer, 2020: 51-67.
[16]	Zhang Jinming, Hu Xiangyun, Wei Yujun, et al. Road topology extraction from satellite imagery by joint learning of nodes and their connectivity[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 5602613.
[17]	Bastani F, He Songtao, Abbar S, et al. RoadTracer: automatic extraction of road networks from aerial images[C]//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 4720-4728.
[18]	Xu Zhenhua, Liu Yuxuan, Gan Lu, et al. RNGDet: road network graph detection by transformer in aerial images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 4707612.
[19]	王泽矫, 向隆刚, 王猛, 等. 融合层级特征与多样化注意力的道路面与中心线协同提取网络[J]. 测绘学报, 2026, 55(3): 548-563. DOI: . doi: 10.11947/j.AGCS.2026.20250446
	Wang Zejiao, Xiang Longgang, Wang Meng, et al. Hierarchical feature and diversified attention fusion network for collaborative extraction of road surface and centerline[J]. Acta Geodaetica et Cartographica Sinica, 2026, 55(3): 548-563. DOI: . doi: 10.11947/j.AGCS.2026.20250446
[20]	方志祥, 仲浩宇, 邹欣妍. 轨迹延续性与影像特征相似性结合的城市道路提取[J]. 测绘学报, 2020, 49(12): 1554-1563. DOI: . doi: 10.11947/j.AGCS.2020.20190366
	Fang Zhixiang, Zhong Haoyu, Zou Xinyan. Extracting urban road area based on combination of trajectory continuity and image feature similarity[J]. Acta Geodaetica et Cartographica Sinica, 2020, 49(12): 1554-1563. DOI: . doi: 10.11947/j.AGCS.2020.20190366
[21]	Yang Xue, Fan Xiang, Su Yichun, et al. TR2RM: an urban road network generation model based on multisource big data[J]. International Journal of Digital Earth, 2024, 17(1): 2344596.
[22]	李雅丽. 轨迹数据与遥感影像多层次融合的城市路网结构提取[D]. 武汉: 武汉大学, 2022: 76-78.
	Li Yali. Extraction of urban road network structure by multi-level fusion of trajectory data and remote sensing images[D]. Wuhan: Wuhan University, 2022: 76-78.
[23]	Liu Lingbo, Yang Zewei, Li Guanbin, et al. Aerial images meet crowdsourced trajectories: a new approach to robust road extraction[J]. IEEE Transactions on Neural Networks and Learning Systems, 2023, 34(7): 3308-3322.
[24]	Zhang Chong, Zhang Huazu, Guo Xiaogang, et al. DF-BSFNet: a bilateral synergistic fusion network with novel dynamic flow convolution for robust road extraction[J]. Information Fusion, 2025, 118: 102958.
[25]	邓敏, 罗斌, 唐建波, 等. 顾及轨迹密度分布异质性的道路交叉口提取方法[J]. 测绘学报, 2023, 52(6): 1000-1009. DOI: . doi: 10.11947/j.AGCS.2023.20210173
	Deng Min, Luo Bin, Tang Jianbo, et al. Extracting road intersections from vehicle trajectory data in the face of trace density disparity[J]. Acta Geodaetica et Cartographica Sinica, 2023, 52(6): 1000-1009. DOI: . doi: 10.11947/j.AGCS.2023.20210173
[26]	王序哲. 局部自适应二值化方法研究[J]. 软件导刊, 2011, 10(11): 13-14.
	Wang Xuzhe. Research on local adaptive binarization method[J]. Software Guide, 2011, 10(11): 13-14.
[27]	Hetang C, Xue H, Le C, et al. Segment anything model for road network graph extraction[C]//Proceedings of 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Seattle: IEEE, 2024: 2556-2566.
[28]	Luo An, Chen Shenghua, Xu Bin. Enhanced map-matching algorithm with a hidden Markov model for mobile phone positioning[J]. ISPRS International Journal of Geo-Information, 2017, 6(11): 327.
[29]	Biagioni J, Eriksson J. Inferring road maps from global positioning system traces: survey and comparative evaluation[J]. Transportation Research Record: Journal of the Transportation Research Board, 2012, 2291(1): 61-71.

立交桥	精确率	召回率	GEO-F₁值	TOPO-F₁值
丽泽桥	0.939 8	0.905 1	0.922 1	0.934 7
健翔桥	0.971 9	0.922 5	0.946 5	0.952 5
建国门桥	0.901 3	0.847 5	0.873 6	0.896 6
四元桥	0.888 4	0.891 4	0.889 9	0.906 8
新兴桥	0.918 9	0.905 8	0.912 3	0.908 7
莲花桥	0.956 7	0.938 6	0.947 6	0.957 5
安慧桥	0.978 8	0.973 9	0.976 3	0.980 4
定慧桥	0.890 2	0.925 3	0.907 4	0.927 2

判别阈值τ	对应夹角l/（°）	GEO-F₁值	TOPO-F₁值
0.75	138.6	0.878 5	0.895 6
0.85	148.2	0.889 9	0.906 8
0.95	161.8	0.886 7	0.899 4

立交桥	无影像候选点补充		无拓扑纠正		无地图匹配修正
立交桥	GEO-F₁值	TOPO-F₁值	GEO-F₁值	TOPO-F₁值	GEO-F₁值	TOPO-F₁值
丽泽桥	0.866 3	0.881 3	0.886 0	0.897 9	0.913 4	0.922 2
健翔桥	0.935 5	0.940 0	0.925 3	0.931 5	0.944 8	0.948 5
四元桥	0.825 9	0.858 2	0.859 8	0.890 1	0.864 6	0.891 2
新兴桥	0.872 9	0.866 8	0.889 5	0.883 4	0.905 1	0.901 6
安慧桥	0.894 8	0.909 8	0.967 6	0.978 8	0.966 2	0.974 2

立交桥	方法	精确率	召回率	GEO-F₁值	TOPO-F₁值
丽泽桥	RoadRunner	0.898 2	0.509 7	0.650 4	0.681 7
	TrackFusion	0.977 6	0.878 3	0.925 3	0.925 8
	Sat2Graph	0.389 6	0.239 8	0.296 9	0.443 7
	OSM	0.969 1	0.903 4	0.935 1	0.946 7
	本文方法	0.939 8	0.905 1	0.922 1	0.934 7
健翔桥	RoadRunner	0.949 7	0.565 4	0.708 8	0.719 6
	TrackFusion	0.969 0	0.864 6	0.913 9	0.922 1
	Sat2Graph	0.681 2	0.278 2	0.395 1	0.446 2
	OSM	0.969 1	0.956 1	0.962 6	0.971 0
	本文方法	0.971 9	0.922 5	0.946 5	0.952 5
四元桥	RoadRunner	0.787 3	0.600 0	0.681 0	0.719 0
	TrackFusion	0.860 2	0.816 1	0.837 6	0.849 5
	Sat2Graph	0.358 9	0.311 4	0.333 5	0.458 3
	OSM	0.935 8	0.826 9	0.878 0	0.886 3
	本文方法	0.888 4	0.891 4	0.889 9	0.906 8
新兴桥	RoadRunner	0.981 1	0.745 5	0.847 2	0.859 1
	TrackFusion	0.971 2	0.742 2	0.841 4	0.853 5
	Sat2Graph	0.424 9	0.329 7	0.371 3	0.555 6
	OSM	0.977 7	0.880 8	0.926 7	0.891 2
	本文方法	0.918 9	0.905 8	0.912 3	0.908 7
安慧桥	RoadRunner	0.970 1	0.773 1	0.860 4	0.873 5
	TrackFusion	0.961 0	0.805 4	0.876 3	0.887 4
	Sat2Graph	0.527 7	0.150 4	0.234 1	0.326 3
	OSM	0.977 5	0.974 5	0.976 0	0.975 2
	本文方法	0.978 8	0.973 9	0.976 3	0.980 4