Land use classification from evidence photos via the integration of vision foundation models and graph neural networks

doi:10.11947/j.AGCS.2026.20260730

Abstract

Abstract:

Under the “Internet plus verification” framework of national land surveys, the review of land use categories based on evidence photos still heavily relies on manual visual interpretation, resulting in low efficiency and strong subjectivity. Land use categories are determined by functional attributes and spatial organization patterns, while evidence photos of different categories often exhibit similar visual characteristics. Therefore, methods relying solely on visual features remain insufficient for fine-grained land use classification. To address this issue, this paper proposes a land use classification method integrating vision foundation models and graph neural networks. The proposed method employs vision foundation models to extract instance-level semantic features of ground objects, and utilizes multi-view photo-based 3D reconstruction to map cross-view instances into a unified spatial coordinate system. A graph-based representation integrating semantic information and three-dimensional spatial relationships is then constructed to jointly characterize parcel-scale information, and a graph neural network is further employed to perform land use classification. To evaluate the effectiveness of the proposed method, a dedicated evidence-photo dataset for fine-grained construction land categories was constructed from real-world land change verification tasks. Experimental results demonstrate that, compared with the best-performing feature-level fusion ResNet-50 baseline, the proposed method improves the overall accuracy by 9.87 percentage points and increases the macro-F1 score by 11.46 percentage points. In particular, the F1 score of retail commercial land is improved from 25.84% to 61.41%.

Key words: land use classification, evidence photos, vision foundation model, graph neural network

CLC Number:

P237

Jianmei Wang, Yu Duan, Shaoming Zhang, Xinyan Li. Land use classification from evidence photos via the integration of vision foundation models and graph neural networks[J]. Acta Geodaetica et Cartographica Sinica, 2026, 55(8): 1400-1413.

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

URL: http://xb.chinasmp.com/EN/10.11947/j.AGCS.2026.20260730

http://xb.chinasmp.com/EN/Y2026/V55/I8/1400

Figures/Tables 15

Fig. 1

Fig. 2

Tab. 1

Fig. 3

Tab. 2

Tab. 3

Tab. 4

Tab. 5

Fig. 4

Fig. 5

Fig. 6

Fig. 7

Tab. 6

Tab. 7

Tab. 8

References 29

[1]	中国国家标准化管理委员会. 土地利用现状分类：GB/T 21010—2017[S]. 北京: 中国标准出版社, 2017.
	Standardization Administration of China. Current land use classification: GB/T 21010—2017[S]. Beijing: Standards Pressof China, 2017.
[2]	吴田军, 李曼嘉, 骆剑承, 等. 耦合空间分布模式的复杂山区地块作物遥感分类方法[J]. 测绘学报, 2025, 54(7): 1215-1229. DOI: . doi: 10.11947/j.AGCS.2025.20240440
	Wu Tianjun, Li Manjia, Luo Jiancheng, et al. Farmland-parcel-based crop remote sensing classification method in com-plex mountainous areas via coupling spatial distribution patterns[J]. Acta Geodaetica et Cartographica Sinica, 2025, 54(7): 1215-1229. DOI: . doi: 10.11947/j.AGCS.2025.20240440
[3]	孙一帆, 刘冰, 余旭初, 等. 图像级高光谱影像高分辨率特征网络分类方法[J]. 测绘学报, 2024, 53(1): 50-64. DOI: . doi: 10.11947/j.AGCS.2024.20220058
	Sun Yifan, Liu Bing, Yu Xuchu, et al. A high-resolution feature network image-level classification method for hyper-spectral image[J]. Acta Geodaetica et Cartographica Sinica, 2024, 53(1): 50-64. DOI: . doi: 10.11947/j.AGCS.2024.20220058
[4]	胡鑫, 王心宇, 钟燕飞. 基于自适应上下文聚合网络的双高遥感影像分类[J]. 测绘学报, 2023, 52(7): 1175-1186. DOI: . doi: 10.11947/j.AGCS.2023.20220237
	Hu Xin, Wang Xinyu, Zhong Yanfei. Adaptive context aggregation network for H² remote sensing imagery classification[J]. Acta Geodaetica et Cartographica Sinica, 2023, 52(7): 1175-1186. DOI: . doi: 10.11947/j.AGCS.2023.20220237
[5]	赵一鸣, 胡克林, 涂可龙, 等. 基于SAR与光学遥感影像融合的多标签场景分类方法[J]. 测绘学报, 2025, 54(5): 911-923. DOI: . doi: 10.11947/j.AGCS.2025.20240281
	Zhao Yiming, Hu Kelin, Tu Kelong, et al. Multi-label scene classification method based on fusion of SAR and optical remote sensing images[J]. Acta Geodaetica et Cartographica Sinica, 2025, 54(5): 911-923. DOI: . doi: 10.11947/j.AGCS.2025.20240281
[6]	Liu Zhen. Identifying urban land use social functional units: a case study using OSM data[J]. International Journal of Digital Earth, 2021, 14(12): 1798-1817.
[7]	Biljecki F, Ito K. Street view imagery in urban analytics and GIS: a review[J]. Landscape and Urban Planning, 2021, 215: 104217.
[8]	李佳铃, 齐霁, 鲁伟鹏, 等. 面向城市功能区分类的光学遥感影像-OSM数据联合自监督学习方法[J]. 测绘学报, 2025, 54(1): 154-164. DOI: . doi: 10.11947/j.AGCS.2025.20240067
	Li Jialing, Qi Ji, Lu Weipeng, et al. Self-supervised learning based urban functional zone classification by integrating optical remote sensing image-OSM data[J]. Acta Geodaetica et Cartographica Sinica, 2025, 54(1): 154-164. DOI: . doi: 10.11947/j.AGCS.2025.20240067
[9]	Zhao Kun, Li Juan, Xie Shuai, et al. Self-supervised learning with trilateral redundancy reduction for urban functional zone identification using street-view imagery[J]. Sensors, 2025, 25(5): 1504.
[10]	Huang Weiming, Wang Jing, Cong Gao. Zero-shot urban function inference with street view images through prompting a pretrained vision-language model[J]. International Journal of Geographical Information Science, 2024, 38(7): 1414-1442.
[11]	Li Hongsheng, Zhu Guangming, Zhang Liang, et al. Scene graph generation: a comprehensive survey[J]. Neurocomputing, 2024, 566: 127052.
[12]	Li Ziming, Chen Bin, Wu Shengbiao, et al. Deep learning for urban land use category classification: a review and experimental assessment[J]. Remote Sensing of Environment, 2024, 311: 114290.
[13]	Zhang Yan, Li Yong, Zhang Fan. Multi-level urban street representation with street-view imagery and hybrid semantic graph[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2024, 218: 19-32.
[14]	Yang Jianwei, Lu Jiasen, Lee S, et al. Graph R-CNN for scene Graph Generation[C]//Proceedings of 2018 Computer Vision (ECCV 2018). Cham: Springer, 2018: 690-706.
[15]	Chen Gongwei, Song Xinhang, Zeng Haitao, et al. Scene recognition with prototype-agnostic scene layout[J]. IEEE Transactions on Image Processing, 2020, 29: 5877-5888.
[16]	Zeng Haitao, Song Xinhang, Chen Gongwei, et al. Amorphous region context modeling for scene recognition[J]. IEEE Transactions on Multimedia, 2022, 24: 141-151.
[17]	Liang Jiali, Deng Yufan, Zeng Dan. A deep neural network combined CNN and GCN for remote sensing scene classification[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 13: 4325-4338.
[18]	Peng Jingquan, Sun Xian, Yu Hongfeng, et al. An instance-based multitask graph network for complex facility recognition in remote sensing imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 5615015.
[19]	Zhang Yan, Liu Pengyuan, Biljecki F. Knowledge and topology: a two layer spatially dependent graph neural networks to identify urban functions with time-series street view image[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2023, 198: 153-168.
[20]	Kong Bo, Ai Tinghua, Zou Xinyan, et al. A graph-based neural network approach to integrate multi-source data for urban building function classification[J]. Computers, Environment and Urban Systems, 2024, 110: 102094.
[21]	Krishna R, Zhu Yuke, Groth O, et al. Visual genome: connecting language and vision using crowdsourced dense image annotations[J]. International Journal of Computer Vision, 2017, 123(1): 32-73.
[22]	Li Weijia, Yu Jinhua, Chen Dairong, et al. Fine-grained building function recognition with street-view images and GIS map data via geometry-aware semi-supervised learning[J]. International Journal of Applied Earth Observation and Geoinformation, 2025, 137: 104386.
[23]	Dietterich T G, Lathrop R H, Lozano-Pérez T. Solving the multiple instance problem with axis-parallel rectangles[J]. Artificial Intelligence, 1997, 89(1/2): 31-71.
[24]	Ilse M, Tomczak J M, Welling M. Attention-based deep multiple instance learning[C]//Proceedings of the 35th International Conference on Machine Learning. Stockholm: PMLR, 2018, 80: 2127-2136.
[25]	Siméoni O, Vo H V, Seitzer M, et al. DINOv3[PP/OL]. arXiv (2025-08-13) [2026-01-10]. https://doi.org/10.48550/arXiv.2508.10104.
[26]	Kirillov A, Mintun E, Ravi N, et al. Segment anything[C]//Proceedings of 2023 IEEE/CVF International Conference on Computer Vision. Paris: IEEE, 2023: 4015-4026.
[27]	Wang Jianyuan, Chen Minghao, Karaev N, et al. VGGT: visual geometry grounded transformer[C]//Proceedings of 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2025: 5294-5306.
[28]	Kipf T N, Welling M. Semi-supervised classification with graph convolutional networks[C]//Proceedings of the 5th International Conference on Learning Representations. Toulon: IEEE, 2017.
[29]	Cheng Bowen, Misra I, Schwing A G, et al. Masked-attention mask transformer for universal image segmentation[C]//Proceedings of 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans: IEEE, 2022: 1290-1299.

一级建设用地	二级建设用地	含义
商服用地	零售商业用地	以零售功能为主的商铺和加油加气、充换电站等的用地
	商务金融用地	指商务服务用地，以及经营性的办公场所用地
	娱乐用地	指剧院、音乐厅、电影院、歌舞厅、网吧、影视城以及绿地率<65%的大型游乐等设施用地
	其他商业服务业用地	指其他商服用地，含洗浴场所、废旧物资回收站，机动车、电子产品维修网点、物流营业网点
工矿仓储用地	仓储用地	指用于物资储备、中转的场所用地。包括物流仓储设施、配送中心、转运中心等
公共管理与公共服务用地	公园与绿地	指城镇、村庄范围内的公园和用于休憩、美化环境及防护的绿化用地
交通运输用地	城镇村道路用地	指城镇、村庄范围内公用道路及行道树用地
水域及水利设施用地	河流水面	指天然形成或人工开挖河流常水位岸线之间的水面
水域及水利设施用地	沟渠	指人工修建用于引、排、灌的渠道
其他土地	空闲地	指城镇、村庄、工矿范围内尚未使用的土地

隐藏维度	OA
隐藏维度	L=2	L=3	L=4	L=5
H=256	75.5	76.8	75.1	74.4
H=512	74.5	75.8	75.1	73.5
H=1024	74.4	75.1	73.0	71.1
H=2048	65.1	73.3	66.5	70.4

方法	OA	MP	MR	MF₁
决策级融合ResNet-50	61.86	61.02	62.82	60.51
决策级融合ViT-B/16	60.71	60.23	61.75	59.34
特征级融合ResNet-50	63.56	62.58	64.98	62.27
特征级融合ViT-B/16	62.34	61.62	63.95	61.23
本文方法	73.43	73.75	73.88	73.73

土地利用类别	特征级融合ResNet-50			本文方法
土地利用类别	精确率	召回率	F₁值	精确率	召回率	F₁值
零售商业用地	33.82	20.91	25.84	62.65	60.23	61.41
商务金融用地	45.18	74.55	56.26	67.49	74.55	70.84
娱乐用地	36.46	79.55	50.00	82.24	80.00	81.11
其他商服用地	64.97	46.36	54.11	55.26	60.91	57.95
仓储用地	61.34	58.40	59.84	76.16	69.00	72.40
公园与绿地	82.35	85.91	84.09	85.24	87.95	86.58
城镇村道路用地	65.59	59.77	62.54	76.77	71.36	73.97
河流水面	77.85	77.50	77.68	81.65	78.86	80.23
沟渠	71.29	66.59	68.86	74.49	74.32	74.40
空闲地	86.95	80.23	83.45	75.58	81.59	78.47

DINOv3	SAM	VGGT+GCN	OA	MF₁
√	×	×	67.63	67.05
×	√	×	13.82	5.72
√	√	×	68.65	68.40
√	×	√	69.57	69.48
×	√	√	20.68	18.00
√	√	√	73.43	73.73