A domain-adaptive building extraction method from remote sensing imagery using shape and position priors

Zhao Shaoxuan; Zhou Xiaoguang; Hou Dongyang; Guo Jing; Liu Huaguang

doi:10.11834/jrs.20255067

Views : 0 下载量: 0 CSCD: 0

R-PDF
PDF
Export
Share
Collection
Album

A domain-adaptive building extraction method from remote sensing imagery using shape and position priors
Pages: 1-17(2025)
Received：06 January 2009，

Published Online：28 October 2025，
DOI： 10.11834/jrs.20255067
稿件说明：

移动端阅览

赵绍轩，周晓光，侯东阳，郭静，刘华光.XXXX.形状与位置先验信息引导的遥感影像建筑物域适应提取方法.遥感学报，XX（XX）： 1-17 DOI： 10.11834/jrs.20255067.

Zhao Shaoxuan，Zhou Xiaoguang，Hou Dongyang，Guo Jing，Liu Huaguang. XXXX. A domain-adaptive building extraction method from remote sensing imagery using shape and position priors. National Remote Sensing Bulletin， XX（XX）：1-17 DOI： 10.11834/jrs.20255067.

摘要

深度学习为遥感影像建筑物提取提供了一种有效途径，但在训练数据（源域）与测试数据（目标域）分布不同的情况下，仅利用源域训练的模型直接应用在目标域时精度会显著下降。建筑物域适应提取方法可以克服域间数据分布差异，提升不同环境下建筑物信息的识别能力。针对建筑物域适应提取中目标域标签获取难度大、标注成本高，且当前方法未能充分利用建筑物不变特性提供跨域一致性约束的问题，本文提出了形状与位置先验信息引导的遥感影像建筑物域适应提取方法。首先，综合利用建筑物指数、Harris等方法自动提取目标域建筑物角点，并借助图像形态学方法提取源域标签建筑物边缘，作为两域的形状先验信息。然后，设计基于高斯转换的位置先验信息提取方法，将目标域OSM对象和源域标签对象分别转为双域的位置先验信息。最后，利用上述形状先验信息构建形状损失函数，同时在两域对建筑物目标提供训练约束，将双域位置先验信息作为独立附加通道叠合影像图层构成4通道输入，丰富目标域建筑物信息，设计基于对抗学习的建筑物域适应提取模型AU_AdaptNet。实验结果表明：本文方法的IoU指标比未经过域适应的基础模型提高了15%，比未添加先验信息引导的域适应模型提高了6%，且在目标域无OSM数据时，仅依靠形状先验信息引导也能提高建筑物域适应提取精度，在目标域OSM数据完整性较低的条件下，也可与使用高质量目标域标签的半监督域适应方法取得相近效果。

Abstract

Deep learning provides an effective approach for building extraction from remote sensing images. However

when the distribution of train data(source domain) and test data(target domain) is different

the accuracy of a model trained only in the source domain will significantly decrease when applied directly in the target domain. The domain adaptation method can overcome the differences in data distribution between domains

thereby improving the recognition ability of building information in different environments. Therefore

it improves the efficiency of urban planning

post disaster reconstruction

and land use management. The difficulty and cost of obtaining target domain labels in building domain adaptation extraction are high.The current methods fail to fully utilize the invariant properties of buildings to provide cross domain consistency constraints. This paper proposes a domain-adaptive building extraction method from remote sensing imagery guided by shape and position prior information. There are several main components to our approach. First

building corners in the target domain are automatically extracted through the comprehensive application of methods such as the Building Index and Harris operator

while the edges of labeled buildings in the source domain are obtained using image morphological processing. These extracted features serve as shape priors for both domains. Then

we design a Gaussian transformation-based method to convert the OSM objects in the target domain and the labeled objects in the source domain to the position prior information in the dual domain

respectively. Finally

a shape loss is constructed using the aforementioned shape prior information

while training constraints are provided for building in both domains. The dual domain position prior information is used as independent additional channels to superimpose image layers to form a 4-channel input

enriching the building information in the target domain. A building domain adaptation extraction model AU_adaptNet based on adversarial learning is designed.Experimental results show that the IoU metrics of our method can be improved by 15% over the results of the base generative model without domain adaptation

and by 6% over the results of the domain adaptation model without added prior information guidance. When there is no OSM data in the target domain

the domain adaptation accuracy can also be improved by relying only on shape prior information guidance. Under conditions of low OSM data integrity in the target domain

comparable results can be achieved with semi-supervised domain adaptation methods using high-quality target domain labels.We can draw the following conclusions (1) Our method extracts invariant features (e.g.

corners and edges) from source and target domain buildings as shape prior

while simultaneously incorporating crowdsourced data from the target domain as position prior. This leads to the development of a building extraction framework guided by both shape and position prior for domain adaptation. (2) Experimental results demonstrate that when OSM data in target domain exhibits low completeness

the proposed method yields comparable performance to semi-supervised domain adaptation approaches that utilize high-quality target domain labels.

关键词

Keywords

references

Basak H and Yin Z . 2023 . Semi-supervised domain adaptive medical image segmentation through consistency regularized disentangled contrastive learning. International Conference on Medical Image Computing and Computer-Assisted Intervention . Cham : Springer Nature Switzerland , 2023 : 260 - 270 . [ DOI: 10.1007/978-3-031-43901-8_25 http://dx.doi.org/10.1007/978-3-031-43901-8_25 ]

Canny J . 1986 . A Computational Approach to Edge Detection . IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-8 ( 6 ): 679 - 98 . [ DOI: 10.1109/TPAMI.1986.4767851 http://dx.doi.org/10.1109/TPAMI.1986.4767851 ]

Cao Y X and Huang X . 2023 . A full-level fused cross-task transfer learning method for building change detection using noise-robust pretrained networks on crowdsourced labels . Remote Sensing of Environment , 284 : 113371 . [ DOI: 10.1016/j.rse.2022.113371 http://dx.doi.org/10.1016/j.rse.2022.113371 ]

Chen J W , Li Y , Chen J and Chen H Q . 2024 . Building extraction method based on MFF-Deeplabv3+ network for high-resolution remote sensing images . Journal of University of Chinese Academy of Sciences , 41 ( 05 ): 654 - 664

陈经纬 , 李宇 , 陈俊 , 张洪群 . 2024 . 基于MFF-Deeplabv3+网络的高分辨率遥感影像建筑物提取方法 . 中国科学院大学学报(中英文) , 41 ( 05 ): 654 - 664 [ DOI: 10.7523/j. ucas.2023.010 http://dx.doi.org/10.7523/j.ucas.2023.010 ]

Cote M and Saeedi P . 2013 . Automatic rooftop extraction in nadir aerial imagery of suburban regions using corners and variational level set evolution . IEEE transactions on geoscience and remote sensing , 51 ( 1 ): 313 - 328 . [ DOI: 10.1109/TGRS.2012.2200689 http://dx.doi.org/10.1109/TGRS.2012.2200689 ]

Du P J . 2020 . Modern Surveying and Mapping . Progress of High Resolution Remotely Sensed lmage Progressing and Urban Application Examples , 43 ( 01 ): 1 - 9

杜培军 . 2020 . 高分辨率遥感影像处理进展与城市应用若干实例 . 现代测绘 , 43 ( 01 ): 1 - 9

Fan C N , Liu P , Xiao T , Zhao W and Tang X L . 2021 . A Review of Deep Domain Adaptation: General Situation and Complex Situation . Acta Automatica Sinica , 47 ( 03 ): 515 - 548

范苍宁 , 刘鹏 , 肖婷 , 赵巍 , 唐降龙 . 2021 . 深度域适应综述:一般情况与复杂情况 . 自动化学报 , 47 ( 03 ): 515 - 548 [ DOI: 10.16383/j.aas.c200238 http://dx.doi.org/10.16383/j.aas.c200238 ]

Harris C and Stephens M . 1988 . A combined corner and edge detector . Alvey vision conference , 15 ( 50 ): 10 - 5244 .

Huang X and Zhang L P . 2011 . A multidirectional and multiscale morphological index for automatic building extraction from multispectral GeoEye-1 imagery . Photogrammetric engineering and remote sensing , 77 ( 7 ): 721 - 32 . [ DOI: 10.14358/PERS.77.7.721 http://dx.doi.org/10.14358/PERS.77.7.721 ]

Ji S P , Wei S Q and Lu M . 2019 . Fully convolutional networks for multisource building extraction from an open aerial and satellite imagery data set . IEEE Transactions on Geoscience and Remote Sensing , 57 ( 1 ): 574 - 586 . [ DOI: 10.1109/TGRS.2018.2858817 http://dx.doi.org/10.1109/TGRS.2018.2858817 ]

Kaiser P , Wegner J D , Lucchi A , Jaggi M , Hofmann T and Schindler K . 2017 . Learning aerial image segmentation from online maps . IEEE Transactions on Geoscience and Remote Sensing , 55 ( 11 ): 6054 - 6068 . [ DOI: 10.1109/TGRS.2017.2719738 http://dx.doi.org/10.1109/TGRS.2017.2719738 ]

Kang L , Wang Q and Yan H W . 2018 . Building extraction based on OpenStreetMap tags and very high spatial resolution image in urban area . The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences , 42 ( 3 ): 715 - 718 . [ DOI: 10.5194/isprs-archives-XLII-3-715-2018 http://dx.doi.org/10.5194/isprs-archives-XLII-3-715-2018 ]

Li D R . 2018 . Brain Cognition and Spatial Cognition: On Integration of Geo-spatial Big Data and Artificial Intelligence . Geomatics and Information Science of Wuhan University , 43 ( 12 ): 1761 - 1767

李德仁 . 2018 . 脑认知与空间认知——论空间大数据与人工智能的集成 . 武汉大学学报(信息科学版) , 43 ( 12 ): 1761 - 1767 [ DOI: 10.13203/j.whugis20180411 http://dx.doi.org/10.13203/j.whugis20180411 ]

Li Z , Sui Z W , Fu Q Y , Zheng J J and Bu T . 2023 . High-resolution remote sensing extraction of urban buildings based on morphological sequences and multi-source a priori information . National Remote Sensing Bulletin , 27 ( 4 ): 998 - 1008

李治 , 隋正伟 , 傅俏燕 , 卜桐 . 2023 . 基于形态学序列和多源先验信息的城市建筑物高分遥感提取 . 遥感学报 , 27 ( 04 ): 998 - 1008 . [ DOI： 10.11834/jrs.20221077 http://dx.doi.org/10.11834/jrs.20221077 ]

Liu F T , Ting K M and Zhou Z H . 2012 . Isolation-Based Anomaly Detection . ACM Trans Knowl Discovery Data , 6 ( 1 ): 1 - 39 . [ DOI： 10.1145/2133360.2133363 http://dx.doi.org/10.1145/2133360.2133363 ]

Liu H B , Qu Y J , Yan Q S and Deng F . 2023 . The method of variable scale mesh primitive extraction for monomeric and structural reconstruction of building . Acta Geodaetica et Cartographica Sinic , 52 ( 11 ): 1953 - 1961

刘海兵 , 曲英杰 , 颜青松 , 邓非 . 2023 . 建筑物单体结构化重建的变尺度网格基元提取法 . 测绘学报 , 52 ( 11 ): 1953 - 1961 [ DOI: 10.11947/j.AGCS.2023.20220275 http://dx.doi.org/10.11947/j.AGCS.2023.20220275 ]

Liu Y , Zhang Y J , Ao Y , Jiang D L and Zhang Z R . 2024 . Building extraction from remote sensing images via the multiscale information fusion method under the Transformer architecture . National Remote Sensing Bulletin , 28 ( 12 ): 3173 - 3183 .

刘异 , 张寅捷 , 敖洋 , 江大龙 , 张肇睿 . 2024 . Transformer架构下跨多尺度信息融合的遥感影像建筑提取 . 遥感学报 , 28 ( 12 ): 3173 - 3183 . [ DOI: 10.11834/jrs.20233017 http://dx.doi.org/10.11834/jrs.20233017 ]

Long L H , Zhu Y T , Yan J W , Liu J J and Wang Z Y . 2023 . New building extraction method based on semantic segmentation . National Remote Sensing Bulletin ， 27 ( 11 )： 2593 - 2602

龙丽红 , 朱宇霆 , 闫敬文 , 刘敬瑾 , 王宗跃 . 2023 . 新型语义分割D-UNet的建筑物提取 . 遥感学报 , 27 ( 11 ): 2593 - 2602 [ DOI: 10.11834/jrs.20211029 http://dx.doi.org/10.11834/jrs.20211029 ]

Maggiori E , Tarabalka Y , Charpiat G and Alliez P . 2017 . Can semantic labeling methods generalize to any city? the inria aerial image labeling benchmark . 2017 IEEE International geoscience and remote sensing symposium (IGARSS) , 2017 : 3226 - 3229 . [ DOI: 10.1109/IGARSS.2017.8127684 http://dx.doi.org/10.1109/IGARSS.2017.8127684 ]

Markus Gerke I . 2014 . Use of the Stair Vision Library within the ISPRS 2D Semantic Labeling Benchmark (Vaihingen) .

Mcguinness K and O’connor N . 2010 . A comparative evaluation of interactive segmentation algorithms . Pattern Recognition , 43 ( 2 ): 434 - 444 . [ DOI: 10.1016/j.patcog.2009.03.008 http://dx.doi.org/10.1016/j.patcog.2009.03.008 ]

Na Y , Kim J H , Lee K , Park J , Hwang J Y and Choi J . 2021 . Domain adaptive transfer attack-based segmentation networks for building extraction from aerial images . IEEE Transactions on Geoscience and Remote Sensing , 59 ( 6 ): 5171 - 5182 . [ DOI: 10.1109/TGRS.2020.3010055 http://dx.doi.org/10.1109/TGRS.2020.3010055 ]

Peng D F , Guan H Y , Zang Y F and Bruzzone L . 2021 . Full-level domain adaptation for building extraction in very-high-resolution optical remote-sensing images . IEEE Transactions on Geoscience and Remote Sensing , 60 : 1 - 17 . [ DOI: 10.1109/TGRS.2021.3093004 http://dx.doi.org/10.1109/TGRS.2021.3093004 ]

Shi L K , Wang Z Y , Pan B and Shi Z W . 2021 . An end-to-end network for remote sensing imagery semantic segmentation via joint pixel-and representation-level domain adaptation . IEEE Geoscience and Remote Sensing Letters , 18 ( 11 ): 1896 - 1900 . [ DOI: 10.1109/LGRS.2020.3010591 http://dx.doi.org/10.1109/LGRS.2020.3010591 ]

Tasar O , Giros A , Tarabalka Y , Alliez P and Clerc S . 2021 . DAugNet: Unsupervised, multisource, multitarget, and life-long domain adaptation for semantic segmentation of satellite images . IEEE Transactions on Geoscience and Remote Sensing , 59 ( 2 ): 1067 - 1081 . [ DOI: 10.1109/TGRS.2020.3006161 http://dx.doi.org/10.1109/TGRS.2020.3006161 ]

Tsai Y H , Hung W C , Schulter S , Sohn K , Yang M H and Chandraker M . 2018 . Learning to adapt structured output space for semantic segmentation . Proceedings of the IEEE conference on computer vision and pattern recognition , 2018 : 7472 - 7481 .

Viana C M , Encalada L and Rocha J . 2019 . The value of OpenStreetMap historical contributions as a source of sampling data for multi-temporal land use/cover maps . ISPRS International Journal of Geo-Information , 8 ( 3 ): 116 . [ DOI: 10.3390/ijgi8030116 http://dx.doi.org/10.3390/ijgi8030116 ]

Von Gioi R G , Jakubowicz J , Morel J M and Randall G . 2008 . LSD: A fast line segment detector with a false detection control . IEEE transactions on pattern analysis and machine intelligence , 32 ( 4 ): 722 - 732 . [ DOI: 10.1109/TPAMI.2008.300 http://dx.doi.org/10.1109/TPAMI.2008.300 ]

Wang J , Yang X , Qin X , Ye X and Qin Q M . 2014 . An efficient approach for automatic rectangular building extraction from very high resolution optical satellite imagery . IEEE Geoscience and Remote Sensing Letters , 12 ( 3 ): 487 - 491 . [ DOI: 10.1109/LGRS.2014.2347332 http://dx.doi.org/10.1109/LGRS.2014.2347332 ]

Wei D S and Zhou X G . 2019 . Automatic sampling of remote sensing image change detection samples based on prior information of vector data . Journal of Remote Sensing , 23 ( 3 ): 464 – 475 .

魏东升 , 周晓光 . 2019 . 遥感影像变化检测样本自动抽样 . 遥感学报 , 23 ( 03 ): 464 - 475 . [ DOI: 10.11834/jr20198073 http://dx.doi.org/10.11834/jr20198073 ]

Woo S , Park J , Lee J Y and Kweon I S . 2018 . Cbam: Convolutional block attention module . Proceedings of the European conference on computer vision , 2018 : 3 - 19 .

Yan W , Qian Y , Li Y , Wang C and Yang M . 2024 . SS-ADA: A Semi-Supervised Active Domain Adaptation Framework for Semantic Segmentation. arXiv preprint . arXiv:2407. 12788 . [ DOI: 10.48550/arXiv.2407.12788 http://dx.doi.org/10.48550/arXiv.2407.12788 ]

Yan X F , Yuan X , Yang M , Kong B and Liu P C . 2022 . An adaptive building simplification approach based on shape analysis and representation . Acta Geodaetica et Cartographica Sinic , 51 ( 02 ): 269 - 278 .

晏雄锋 , 袁拓 , 杨敏 , 孔博 , 刘鹏程 . 2022 . 建筑物形状特征分析表达与自适应化简方法 . 测绘学报 , 51 ( 02 ): 269 - 278 [ DOI: 10.11947/j.AGCS.2022.20210302 http://dx.doi.org/10.11947/j.AGCS.2022.20210302 ]

Yang H L , Makkar N , Laverdiere M and Rose A . 2024 . Boundary-Aware Adversarial Learning Domain Adaption and Active Learning for Cross-Sensor Building Extraction . IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , 17 : 9122 - 9138 . [ DOI: 10.1109/JSTARS.2024.3385380 http://dx.doi.org/10.1109/JSTARS.2024.3385380 ]

Yang M W , Zhao L K , Ye L F , Jiang W H and Yang Z . 2024 . A review of convolutional neural networks related methods for building extraction from remote sensing images . Journal of Geo-information Science , 26 ( 6 ): 1500 - 1516 .

杨明旺 , 赵丽科 , 叶林峰 , 蒋华伟 , 杨震 . 2024 . 基于卷积神经网络的遥感影像建筑物提取方法综述 . 地球信息科学学报 , 26 ( 06 ): 1500 - 1516

Yiming T , Tang X and Shang H . 2024 . A shape-aware enhancement Vision Transformer for building extraction from remote sensing imagery . International Journal of Remote Sensing , 45 ( 4 ): 1250 - 1276 . [ DOI: 10.1080/01431161.2024.2307325 http://dx.doi.org/10.1080/01431161.2024.2307325 ]

Yin J C , Wu F , Zhai R J , Qiu Y , Gong X Y and Hang R X . 2024 . Two-stream boundary constraints and relativistic generation adversarial network for building contour regularization . Acta Geodaetica et Cartographica Sinic , 53 ( 07 ): 1444 - 1457

殷吉崇 , 武芳 , 翟仁健 , 邱越 , 巩现勇 , 行瑞星 . 2024 . 面向建筑物轮廓规则化的双路径边界约束与相对论生成对抗网络 . 测绘学报 , 53 ( 07 ): 1444 - 1457 [ DOI: 10.11947/j.AGCS.2024.20230056 http://dx.doi.org/10.11947/j.AGCS.2024.20230056 ]

Yu G , Zhou X , Hou D and Wei D . 2021 . Abnormal crowdsourced data detection using remote sensing image features . The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences , 43 : 215 - 221 . [ DOI: 10.5194/isprs-archives-XLIII-B4-2021-215-2021 http://dx.doi.org/10.5194/isprs-archives-XLIII-B4-2021-215-2021 ]

Zhang Z H , Guo W W , Li M J and Yu W X . 2020 . GIS-supervised building extraction with label noise-adaptive fully convolutional neural network . IEEE Geoscience and Remote Sensing Letters , 17 ( 12 ): 2135 - 2139 . [ DOI: 10.1109/LGRS.2019.2963065 http://dx.doi.org/10.1109/LGRS.2019.2963065 ]

Zheng Q , Chen J , Wang Z Y , Jiang J J and Liang C . 2019 . Deep segmentation domain adaptation network with weighted boundary constraint . IEEE Access , 7 : 93909 - 93918 . [ DOI: 10.1109/ACCESS.2019.2927599 http://dx.doi.org/10.1109/ACCESS.2019.2927599 ]

Zhu J Y , Park T , Isola P and Efros A . 2017 . Unpaired image-to-image translation using cycle-consistent adversarial networks . Proceedings of the IEEE international conference on computer vision , 2017 : 2223 - 2232 .

Sun J Y , Huang Z J , Zhou S G , Xu N , Qian H M and Wang C L . 2017 . Building outline vectorization from highspatial resolution imagery . Journal of Remote Sensing , 21 ( 3 ): 396 – 405 .

孙金彦 , 黄祚继 , 周绍光 , 徐南 , 钱海明 , 王春林 . 2017 . 高分辨率遥感影像中建筑物轮廓信息矢量化 . 遥感学报 , 21 ( 03 ): 396 - 405 . [ DOI: 10.11834/jr20176127 http://dx.doi.org/10.11834/jr20176127 ]

Zuo X L , Shao Z F , Wang J M X and Wang Yu . 2024 . A cross-stage features fusion network for building extraction from remote sensing images . Geo-Spatial Information Science , 2024 : 1 - 15 . [ DOI: 10.1080/10095020.2024.2307922 http://dx.doi.org/10.1080/10095020.2024.2307922 ]

Alert me when the article has been cited

提交

Res_ASPP_UNet++： Building an extraction network from remote sensing imagery combining depthwise separable convolution with atrous spatial pyramid pooling

A two-branch building extraction network integrating main body and edge separation and multi-scale information extraction

Mapping typical floodplain wetlands using an improved D-UNet model

MFBFS： A fine-grained building feature set for high-resolution multispectral remote sensing images