面向遥感影像建筑物提取的边界感知动态原型网络

Boundary-Aware Dynamic Prototype Network for Building Extraction from Remote Sensing Imagery

摘要

针对高分辨率遥感影像建筑物提取中边界模糊、背景干扰和轮廓恢复不完整等问题，提出了一种边界感知动态原型网络（DPB-Net）。该网络首先利用平行交叉条带注意力提取水平和垂直方向的空间信息，以增强建筑物边界区域的表达能力；然后，通过双模式动态原型解码器对建筑物区域特征进行建模，降低复杂背景对分割结果的干扰；最后，结合上下文感知重组解码器和边界感知门控机制，逐步恢复建筑物边界与局部轮廓细节，提高预测结果的边界完整性。在WHU和Inria建筑物数据集上的实验结果表明，该方法在WHU数据集上的交并比达到93.94%，与现有方法相比，95%Hausdorff距离降低近20%，并在区域分割精度与边界恢复质量方面表现出优势；在Inria数据集上，交并比达到84.80%，95%Hausdorff距离为18.720 4 px。该方法能够改善复杂城市场景下建筑物区域分割和边界保持效果。

Abstract

To address blurred boundaries, background interference, and incomplete contour recovery in building extraction from high-resolution remote sensing imagery, a dynamic prototype boundary-aware network (DPB-Net) is proposed. First, a parallel cross-strip attention module is designed to capture horizontal and vertical spatial information, thereby enhancing feature representation along building boundaries. Second, a dual-mode dynamic prototype decoder is introduced to model building-region features and suppress interference from complex backgrounds in the segmentation results. Finally, a context-aware reassembly decoder, together with a boundary-aware gating mechanism, progressively recovers building boundaries and local contour details and thus improves boundary completeness in the predictions. Experiments on the WHU and Inria building datasets show that the proposed method achieves IoU scores of 93.94% and 84.80% on the two datasets, respectively. On the WHU dataset, it reduces the 95th-percentile Hausdorff distance (HD₉₅) by nearly 20% compared with existing methods, demonstrating clear advantages in both region segmentation accuracy and boundary recovery quality; on the Inria dataset, it attains an HD₉₅ of 18.720 4 pixels. The proposed method effectively improves building segmentation and boundary preservation in complex scenes.

关键词

建筑物提取 / 高分辨率遥感影像 / 边界感知 / 动态原型 / 深度学习

Key words

building extraction / high-resolution remote sensing imagery / boundary awareness / dynamic prototype / deep learning

参考文献

[1] 范荣双，陈洋，徐启恒，等. 基于深度学习的高分辨率遥感影像建筑物提取方法[J].测绘学报，2019，48(1)：34-41.

[2] 朱盼盼，李帅朋，张立强，等. 基于多任务学习的高分辨率遥感影像建筑提取[J].地球信息科学学报，2021，23(3)：514-523.

[3] 李传林，黄风华，胡威，等. 基于Res_AttentionUnet的高分辨率遥感影像建筑物提取方法[J].地球信息科学学报，2021，23(12)：2232-2243.

[4] 赵元昊，赵莹莹，刘东升，等. 遥感影像建筑物提取多尺度特征深度学习网络[J].航天返回与遥感，2022，43(4)：25-35.

[5] Wang L B, Fang S H, Meng X L, et al. Building extraction with vision transformer[J].IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 5625711.

[6] Li J P, He W, Cao W N, et al. UANet: An uncertainty-aware network for building extraction from remote sensing images[J].IEEE Transactions on Geoscience and Remote Sensing, 2024, 62: 5608513.

[7] Pang Z J, Hu R M, Zhu W, et al. A building extraction method for high-resolution remote sensing images with multiple attentions and parallel encoders combining enhanced spectral information[J].Sensors, 2024, 24(3): 1006.

[8] Li Y J, Li Y G, Zhu X B, et al. A method for extracting buildings from remote sensing images based on 3DJA-UNet3+[J].Scientific Reports, 2024, 14: 19067.

[9] Wang W L, Yu P, Li M M, et al. TDFNet: Twice decoding V-Mamba-CNN fusion features for building extraction[J].Geo-spatial Information Science, 2026, 29(1): 19-38.

[10] Zhang Y, Deng J Y, Yan W J. Urban building footprints extraction and heights estimation from high-resolution spaceborne remote sensing imagery using a CNN-transformer network [J].Remote Sensing, 2026, 18(10): 1484.

[11] Li F, Wang X, Wang H C, et al. LMVMamba: A hybrid U-shape mamba for remote sensing segmentation with adaptation fine-tuning [J].Remote Sensing, 2025, 17(19): 3367.

[12] 张艳，王翔宇，张众维，等. 一种基于边界感知的遥感影像建筑物提取方法[J].西安电子科技大学学报（自然科学版），2022，49(1)：236-244.

[13] Woo S, Debnath S, Hu R H, et al. ConvNeXt V2: Co-designing and scaling ConvNets with masked autoencoders[C]//CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2023: 16133-16142.

[14] Wang J Q, Chen K, Xu R, et al. CARAFE: content-aware ReAssembly of FEatures[C]//CVF International Conference on Computer Vision (ICCV). Piscataway, NJ: IEEE, 2019: 3007-3016.

[15] Ji S P, Wei S Q, Lu M. Fully convolutional networks for multisource building extraction from an open aerial and satellite imagery data set[J].IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(1): 574-586.

[16] Maggiori E, Tarabalka Y, Charpiat G, et al. Can semantic labeling methods generalize to any city? The inria aerial image labeling benchmark[C]//International Geoscience and Remote Sensing Symposium (IGARSS). Piscataway, NJ: IEEE, 2017: 3226-3229.

[17] Taha A A, Hanbury A. Metrics for evaluating 3D medical image segmentation: Analysis, selection, and tool[J].BMC Medical Imaging, 2015, 15(1): 29.

[18] Ronneberger O, Fischer P, Brox T. U-Net: Convolutional networks for biomedical image segmentation [C]//Medical Image Computing and Computer-Assisted Intervention (MICCAI). Cham: Springer International Publishing, 2015: 234-241.

[19] Zhao H S, Shi J P, Qi X J, et al. Pyramid scene parsing network[C]//Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ: IEEE, 2017: 6230-6239.

[20] Chen L C, Zhu Y K, Papandreou G, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//Computer Vision - ECCV. Cham: Springer International Publishing, 2018: 833-851.

[21] Wang J D, Sun K, Cheng T H, et al. Deep high-resolution representation learning for visual recognition [J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(10): 3349-3364.

[22] Liu Z, Lin Y T, Cao Y, et al. Swin transformer: Hierarchical vision transformer using shifted windows [C]//CVF International Conference on Computer Vision (ICCV). Piscataway, NJ: IEEE, 2021: 9992-10002.

[23] Xie E Z, Wang W H, Yu Z D, et al. SegFormer: Simple and efficient design for semantic segmentation with transformers[C]//Ranzato M, Beygelzimer A, Dauphin Y,et al. Proceedings of the 35th International Conference on Neural Information Processing Systems. New York: Curran Associates Inc., 2021: 12077-12090.

[24] Xiao T T, Liu Y C, Zhou B L, et al. Unified perceptual parsing for scene understanding[C]//Computer Vision - ECCV. Cham: Springer International Publishing, 2018: 432-448.

基金资助

国家自然科学基金项目(62365016)