基于预训练迁移学习的高原区域高分辨率水体提取

High-Resolution Water Body Extraction in Plateau Regions Based on Pre-Trained Transfer Learning

摘要

针对为提升三江源区域高分辨率水体提取的细粒度问题, 通过对吉林一号亚米级遥感影像的研究与分析, 提出一种基于私有数据预训练的迁移学习方法。 基于 ViT(Vision Transformer)骨干网络进行自监督预训练, 并构建以 ViT-B 为编码器、UperNet 为解码器的水体迁移学习提取框架。 实验结果表明, 该方法的 IoU(Intersection over Union)达到 92.93%, 较未预训练模型提升 11.87%, 且优于其他开源预训练权重。 最终提取水体图斑 25 621 个, 水体面积 1 544 km²。 与 WorldCover 2021 对比分析显示, 该方法可有效检测出被遗漏的细小水体, 为高原湿地水资源分析提供了更精准的参考。

Abstract

This study aims to improve the accuracy of high-resolution water body extraction in the Sanjiangyuan region. Using sub-meter Jilin-1 satellite imagery, a self-supervised pretraining approach based on the ViT(Vision Transformer) backbone is developed, followed by fine-tuning a water body extraction model with ViT-B as the encoder and UperNet as the decoder. The results demonstrate that the proposed method achieves an IoU(Intersection over Union) of 92.93%, outperforming non-pretrained models by 11.87% and surpassing other open-source pretrained weights. A total of 25 621 water body patches with an area of 1 544 km ² are extracted. Comparative analysis with WorldCover 2021 reveals that the method effectively detects small water bodies missed by previous methods, providing a more accurate reference for water resource analysis in plateau wetlands.

关键词

高分辨率遥感影像 / 预训练 / 迁移学习 / 三江源地区 / 微小水体提取

Key words

high-resolution remote sensing images / pre-trained models / transfer learning / Sanjiangyuan region / water body extraction

参考文献

[1] 陈静, 王晓轩, 吴宇静, 等. 基于 CNN 的零样本城市遥感影像场景分割算法[J]. 吉林大学学报(信息科学版), 2023, 41(4): 739-745.

[2] CHEN J, WANG X X, WU Y J, et al. Zero-Sample Urban Remote Sensing Image Scene Segmentation Algorithm Based on Convolutional Neural Network[J]. Journal of Jilin University (Information Science Edition), 2023, 41(4): 739-745.

[3] 李王波, 范昕桐, 顾玲嘉. 基于星载被动微波的中国东北森林雪深反演[J]. 吉林大学学报(信息科学版), 2023, 41(5): 914-921.

[4] LI W B, FAN X T, GU L J. Snow Depth Retrieval for Forest Area in Northeast China Based on Spaceborne Passive Microwave[J]. Journal of Jilin University (Information Science Edition), 2023, 41(5): 914-921.

[5] BANGIRA T, ALFIERI S M, MENENTI M, et al. Comparing Thresholding with Machine Learning Classifiers for Mapping Complex Water [J/OL]. Remote Sensing, 2019, 11(11): 1351 [2025-04-20]. https://doi.org/10.3390/rs11111351.

[6] 都金康, 黄永胜, 冯学智, 等. SPOT 卫星影像的水体提取方法及分类研究[J]. 遥感学报, 2001(3): 214-219.

[7] DU J K, HUANG Y S, FENG X Z, et al. Research on Water Extraction Method and Classification of SPOT Satellite Images[J]. Journal of Remote Sensing, 2001(3): 214-219.

[8] 陈文倩, 丁建丽, 李艳华, 等. 基于国产 GF-1 遥感影像的水体提取方法[J]. 资源科学, 2015, 37(6): 1166-1172.

[9] CHEN W Q, DING J L, LI Y H, et al. Water Extraction Method Based on Domestic GF-1 Remote Sensing Images[J]. Resource Science, 2015, 37(6): 1166-1172.

[10] 张晗涛, 胡荣明, 姜友谊, 等. 改进 DeeplabV3+模型的河流水体提取[J]. 遥感信息, 2023, 38(3): 146-152.

[11] ZHANG H T, HU R M, JIANG Y Y, et al. River Water Extraction with Improved DeeplabV3+ Model[J]. Remote Sensing Information, 2023, 38(3): 146-152.

[12] 杨笑天, 鱼昕, 刘铭, 等. 基于 UPCBAM-RYOLO V5 的光学遥感舰船小目标检测[J]. 吉林大学学报(信息科学版), 2024, 42(6): 1048-1057.

[13] YANG X T, YU X, LIU M, et al. Optical Remote Sensing Ship Small Target Detection Based on UPCBAM-RYOLO V5[J]. Journal of Jilin University (Information Science Edition), 2024, 42(6): 1048-1057.

[14] RONNEBERGER O, FISCHER P, BROX T. U-Net: Convolutional Networks for Biomedical Image Segmentation[C]// Medical Image Computing and Computer-Assisted Intervention-MICCAI 2015. Cham: Springer, 2015: 234-241.

[15] ZHAO H, SHI J, QI X, et al. Pyramid Scene Parsing Network[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, NJ, USA: IEEE, 2017: 6230-6239.

[16] CHEN L C, PAPANDREOU G, KOKKINOS I, et al. DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(4): 834-848.

[17] BADRINARAYANAN V, KENDALL A, CIPOLLA R. SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(12): 2481-2495.

[18] LIN G, MILAN A, SHEN C, et al. RefineNet: Multi-Path Refinement Networks for High-Resolution Semantic Segmentation[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, NJ, USA: IEEE, 2017: 1925-1934.

[19] LI J, MENG Y, LI Y, et al. Accurate Water Extraction Using Remote Sensing Imagery Based on Normalized Difference Water Index and Unsupervised Deep Learning [J/OL]. Journal of Hydrology, 2022, 615: 128202 [2025-04-20]. https://doi.org/10.1016/j.jhydrol.2022.128202.

[20] DENG J, DONG W, SOCHER R, et al. ImageNet: A Large-Scale Hierarchical Image Database[C]// 2009 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, NJ, USA: IEEE, 2009: 248-255.

[21] HE K, CHEN X, XIE S, et al. Masked Autoencoders Are Scalable Vision Learners[C]// 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway, NJ, USA: IEEE, 2022: 16000-16009.

[22] OQUAB M, DARCET T, MOUTAKANNI T, et al. DINOv2: Learning Robust Visual Features without Supervision [DB/OL]. (2023-04-14)[2025-03-15]. https://arxiv.org/abs/2304.07193.

[23] DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An Image Is Worth 16x16 Words: Transformers for Image Recognition at Scale [DB/OL]. (2020-10-22)[2025-03-15]. https://arxiv.org/abs/2010.11929.

[24] XIAO T, LIU Y, ZHOU B, et al. Unified Perceptual Parsing for Scene Understanding[C]// Proceedings of the European Conference on Computer Vision. Cham: Springer, 2018: 418-434.

[25] OSGOUEI P E, SERTEL E, KABADAYI M E. Assessing the Accuracy of the Esa Worldcover 2021 for the Local Region of Lalapasa/Edirne, Turkey and Recommending Possible Accuracy Improvement Strategies[C]// 2023 11th International Conference on Agro-Geoinformatics. Piscataway, NJ, USA: IEEE, 2023: 1-4.

[26] HUANG L, YOU S, ZHENG M, et al. Green Hierarchical Vision Transformer for Masked Image Modeling[J]. Advances in Neural Information Processing Systems, 2022, 35: 19997-20010.

基金资助

长春市科技发展计划基金资助项目(2024WX02)