Intelligent Glacial Lake Identification in Complex Plateau Terrain Regions Using Multi-Source Remote Sensing Data and Mask R-CNN Deep Learning Model
-
摘要: 冰湖识别是了解冰湖对气候变化的响应和评估冰湖溃决洪水潜在危险的先决条件. 虽然遥感技术使全球冰湖演变的持续监测和评估成为可能，但准确可靠地提取复杂高原地形区的冰湖仍然具有挑战性.提出了融合多源遥感数据和改进后Mask R-CNN深度学习模型的复杂高原地形区冰湖智能识别方法，在Mask R-CNN模型基础上，通过在骨干网络ResNet-50的高层特征（Conv4和Conv5）、FPN的每个特征图以及Mask Head中引入注意力机制. 利用Sentinel-2高分辨遥感影像、ALOS-DEM及NDWI数据组成多波段数据集，并在青藏高原东南部的林芝市进行测试，并进一步比较了改进后Mask R-CNN、U-Net、SegNet和DeepLab V3模型在冰湖识别中的性能.改进后的Mask R-CNN模型具有更高的准确率，模型的精确度、召回率和准确度值分别达到了91.25%、93.69%、92.89%.它有效地降低了山体阴影、湖水浊度和冻融湖水条件对冰湖识别的影响，并显著提高了小冰湖的识别效率. 为地形复杂高原地形区冰湖识别提供了可靠解决方案，为深度学习与多源遥感数据结合的智能化冰湖提取提供了新的框架和可能性.
-
关键词:
- 冰湖 /
- 复杂高原地形区 /
- 智能识别 /
- Mask R‐CNN模型 /
- 多源遥感数据
Abstract: The identification of glacial lakes is a prerequisite for understanding their response to climate change and assessing potential risks of glacial lake outburst floods (GLOFs). Although remote sensing technology enables continuous monitoring and assessment of global glacial lake evolution, accurately and reliably extracting glacial lakes in complex plateau terrain regions remains challenging. This study proposes an intelligent glacial lake identification method for complex plateau terrain based on multi-source remote sensing data and an improved Mask R-CNN deep learning model. Building upon the original Mask R-CNN framework, we introduce attention mechanisms at three key components: the high-level features (Conv4 and Conv5) of the ResNet-50 backbone network, each feature map in the Feature Pyramid Network (FPN), and the Mask Head. Utilizing a multi-band dataset composed of Sentinel-2 high-resolution imagery, ALOS-DEM, and Normalized Difference Water Index (NDWI) data, we conducted tests in Nyingchi City, southeastern Tibetan Plateau. Comparative analyses were performed between the enhanced Mask R-CNN model and three other models (U-Net, SegNet, and DeepLab V3) for glacial lake identification. Results demonstrate that the improved Mask R-CNN achieves superior accuracy, with precision, recall, and accuracy values reaching 91.25%, 93.69%, and 92.89% respectively. The enhanced model effectively mitigates interference from mountain shadows, lake turbidity, and freeze-thaw conditions on glacial lake identification while significantly improving detection efficiency for small glacial lakes. This research provides a reliable solution for glacial lake identification in complex plateau terrain regions and establishes a novel framework combining deep learning with multi-source remote sensing data for intelligent glacial lake extraction, offering new possibilities for related studies. -
图 2 复杂高原地形区冰湖智能识别流程
Fig. 2. Intelligent recognition process of glacial lakes in complex plateau terrain
图 4 不同模型对冰湖识别的结果对比
Fig. 4. Comparison of results for glacial lake recognition across different models
表 1 数据来源与特点
Table 1. Data Sources and Characteristics
数据 分辨率 来源 Sentinel-2 10.0 m https://dataspace.copernicus.eu/ ALOS-DEM 12.5 m https://search.asf.alaska.edu/ 2017年亚洲高山区30 m分辨率冰湖数据集 / https://doi.org/10.5281/zenodo.4275164 
下载: 导出CSV
表 2 使用影像的详细信息
Table 2. Detailed information on the images used
获取日期 云量(%) 数据等级 用途 获取日期 云量(%) 数据等级 用途 2023-11-01 0.05 L2A 模型训练 2023-11-17 3.50 L2A 模型测试 2023-11-01 0.02 L1C 模型训练 2023-10-13 4.49 L1C 模型测试 2023-10-13 8.65 L1C 模型训练 2023-11-19 0.35 L2A 模型测试 2023-11-19 4.70 L2A 模型训练 2023-11-19 0.35 L2A 模型测试 2023-11-19 2.94 L1C 模型训练 2023-11-19 0.79 L1C 模型测试 2023-11-19 1.68 L1C 模型训练 2023-11-01 1.78 L1C 模型测试 2023-11-19 1.68 L1C 模型训练 2023-10-23 0.03 L1C 模型测试 2023-11-01 0.00 L1C 模型训练 2023-11-17 3.31 L1C 模型测试 2023-11-01 0.27 L2A 模型训练 2023-11-19 0.55 L2A 模型测试 2023-10-29 2.95 L2A 模型训练 2023-11-19 2.13 L2A 模型测试 2023-11-09 0.68 L2A 模型训练 2023-11-19 1.21 L1C 模型测试 2023-11-11 1.91 L1C 模型训练 2023-11-01 1.93 L2A 模型测试 2023-11-01 0.48 L2A 模型训练 2023-11-01 1.94 L2A 模型测试
下载: 导出CSV
表 3 模型精度对比
Table 3. Comparison of Model Accuracy
模型 精确率 召回率 准确度 改进后Mask R-CNN 93.25% 94.69% 94.89% U-Net 87.31% 86.54% 90.29% SegNet 81.71% 84.39% 85.58% DeepLab V3 84.33% 86.53% 86.78%
下载: 导出CSV
-
Bhardwaj, A., Singh, M. K., Joshi, P. K., et al., 2015. A Lake Detection Algorithm (LDA) Using Landsat 8 Data: a Comparative Approach in Glacial Environment. International Journal of Applied Earth Observation and Geoinformation, 38: 150-163. https://doi.org/10.1016/j.jag.2015.01.004 Cao, Y., 2022. Automatic Detection of Himalayan Glacial LakesUsing Deep Convolutional Neural Network (Dissertation). University of Electronic Science and Technology of China, Chengdu (in Chinese with English abstract). Che, Y. J., Chen, L. H., Wu, J. K., et al., 2024. Progress in the Study on The Interaction Between Proglacial Lake and Lake-Terminating Glacier Over the Qinghai-Tibet Plateau. Climate Change Research, 20(5): 519-533 (in Chinese with English abstract). Chen, F., Wang, J. X., Zhang, M. M., et al., 2023. Comparative Study on the Extraction Methods of Himalayan Glacial Lakes Based on Historical Boundaries. Journal of Glaciology and Geocryology, 45(4): 1413-1427 (in Chinese with English abstract). Chen, G. Y., Li, T., Chen, J., et al., 2023. Primary Establishment of an Early Warning Model of Debris Flow Hazards In Nyingchi City of Tibetan Autonomous Region Based on Raster Runoff Simulation. The Chinese Journal of Geological Hazard and Control, 34(1): 110-120 (in Chinese with English abstract). Chen, H. W., 2022. Study on Vulnerability Assessment and Risk Prediction of Glacier Lake Outburst Floods in Tibetan Plateau (Dissertation). Chengdu University of Technology, Chengdu (in Chinese with English abstract). Cheng, S., Li, J. R., Wang, Z. Q., et al., 2024. Lightweight Underwater Optical Image Recognition Algorithm Based on YOLOv8. Laser & Optoelectronics Progress, 62(4): 402-412 (in Chinese with English abstract). Dell, R. L., Banwell, A. F., Willis, I. C., et al., 2022. Supervised Classification of Slush and Ponded Water on Antarctic Ice Shelves Using Landsat 8 Imagery: CORRIGENDUM. Journal of Glaciology, 68(268): 415-416. https://doi.org/10.1017/jog.2022.15. Dong, A. N., Dou, J., Li, C. D., et al., 2024. Accelerating Cross-Scene Co-Seismic Landslide Detection through Progressive Transfer Learning and Lightweight Deep Learning Strategies. IEEE Transactions on Geoscience and Remote Sensing, 62: 4410213. https://doi.org/10.1109/TGRS.2024.3424680 Dou, J., Xing, K., Wang, L. Z., et al., 2025. Air-Space-Ground Synergistic Observations for Rapid Post-Seismic Disaster Assessment of 2025 Ms6.8 Xigaze Earthquake, Xizang. Journal of Earth Science, 1-18. https://doi.org/10.1007/s12583-025-0160-2 Dou, J., Yunus, A. P., Merghadi, A., et al., 2020. Different Sampling Strategies for Predicting Landslide Susceptibilities Are Deemed less Consequential with Deep Learning. Science of the Total Environment, 720: 137320. https://doi.org/10.1016/j.scitotenv.2020.137320 Dou, J., Xiang, Z. L., Xu, Q., et al., 2023. Application and Development Trend of Machine Learning in Landslide Intelligent Disaster Prevention and Mitigation. Earth Science, 48(5): 1657-1674(in Chinese with English abstract). Guo, H. L., Xie, Y. L., Hu, L. F., et al., 2024. Water Body Extraction from Remote Sensing Images Based on Improved HarDNet-MSEG. Journal of Geo-Information Science, 26(7): 1745-1762(in Chinese with English abstract). Hu, J. Y., 2024. Evolution of a Typical Moraine Lake and Simulation of Potential GLOF in Southeastern Tibet (Dissertation). Northwest Normal University, Lanzhou (in Chinese with English abstract). Jiang, Z. Y., 2023. Study on Kyagar Glacier Movement and Its Influence on Glacial Lake Variation Under Climate Change (Dissertation). Xinyang Normal University, Xinyang (in Chinese with English abstract). Jin, S. Z., Dai, H. J., Peng, J., et al., 2022. An Improved Mask R-CNN Method for Weed Segmentation. 2022 IEEE 17th Conference on Industrial Electronics and Applications (ICIEA). December 16-19, 2022, Chengdu, China. IEEE, 1430-1435. https://doi.org/10.1109/ICIEA54703.2022.10006300 Li, J. L., Sheng, Y. W., Luo, J. C., 2011. Automatic Extraction of Himalayan Glacial Lakes with Remote Sensing. Journal of Remote Sensing, 15(1): 29-43(in Chinese with English abstract). Li, M. F., Zheng, J. H., Qian, A. L., et al., 2024. Research on the Extraction Method of Tianshan Glacier Lake Based on Decision Tree. Arid Zone Research, 41(10): 1699-1707 (in Chinese with English abstract). Li, X. E., Liu, Y., Jiang, L. M., et al., 2024. A Review on Remote Sensing Monitoring of Glacial Lake Change and Glacial Lake Outburst Floods in Mountain Glacier Regions. Journal of Geo-Information Science, 26(4): 1019-1039(in Chinese with English abstract). Li, Y. C., Zhang, Jun., Liu, C. L., 2021. Extraction Method of Alpine Small Glacial Lake in Qianhu Mountain Area of Yunnan Province Based on Sentinel-2 Image. Science of Surveying and Mapping, 46(4): 114-120 (in Chinese with English abstract). Liu, B. X., Wang, W., Li, W. P., 2023. A Lake Extraction Method Combining the Object-Oriented Method with Boundary Recognition. Land, 12(3): 545. https://doi.org/10.3390/land12030545 Liu, M., 2020. Glacial Lake Outburst Flood/Debris Flow Disaster Mechanism and Hazards Assessment in Bhote Koshi Basin (Dissertation). University of Chinese Academy of Sciences (Institute of Mountain Hazards and Environment, Chinese Academy of Sciences), Chengdu (in Chinese with English abstract). Liu, S. Q., Li, J. L., Li, R. N., et al., 2024. Mapping and Spatial Distribution Characteristics of Glacial Lakes in Xinjiang Based on Sentinel-2 Imagery. Journal of Glaciology and Geocryology, 46(2): 513-524 (in Chinese with English abstract). Lu, M., 2022. Research on Water Extraction Method of High-Resolution Remote Sensing Images by Using Restricted Labeling (Dissertation). Hunan University, Changsha (in Chinese with English abstract). Ma, J. S., 2022. Spatially and Temporally Resolved Monitoring of Global Glacial Lake Changes Based on Multi-Source Remote Sensing (Dissertation). Nanjing University of Information Science and Technology, Nanjing (in Chinese with English abstract). Su, B. Y., Du, X. P., Mu, H. W., et al., 2024. Coupling Mask R-CNN and Attention Mechanism for Building Extraction and Post-Processing Strategy. Remote Sensing Technology and Application, 39(3): 620-632 (in Chinese with English abstract). Tom, M., Prabha, R., Wu, T. Y., et al., 2020. Ice Monitoring in Swiss Lakes from Optical Satellites and Webcams Using Machine Learning. Remote Sensing, 12(21): 3555. https://doi.org/10.3390/rs12213555 Wang, H. R., Wei, J. W., Gao, X. Y., et al., 2023. Enhanced Mapping of Supraglacial Lakes Through Dual-Attention Deep Neural Network. Proceedings of the 31st ACM International Conference on Advances in Geographic Information Systems, 1-4. https://doi.org/10.1145/3589132. 3629972 doi: 10.1145/3589132.3629972 Wei, C. T., Zhao, W. J., Sun, B. C., et al., 2024. Intelligent Rebar Inspection Based on Improved Mask R-CNN and Stereo Vision. Journal of Zhejiang University (Engineering Science), 58(5): 1009-1019 (in Chinese with English abstract). Yang, C. D., Wang, X., Wei, J. F., et al., 2019. Chinese Glacial Lake Inventory Based on 3S Technology Method. Acta GeographicaSinica, 74(3): 544-556 (in Chinese with English abstract). Yang, N. T., Nie, Y., 2024. An Improved Deep Learning Method for Mapping Glacial Lakes Using Satellite Observation and Its Application. Spacecraft Recovery & Remote Sensing 45(1): 41-52(in Chinese with English abstract). Yin, L. C., Wang, X., Yin, Y. S., et al., 2024. Automatic Extraction of Glacial Lakes Based on Deep Learning and Sentinel-2 Imagery. Remote Sensing Technology and Application, 39(6): 1319-1329 (in Chinese with English abstract). Yu C, Sun Y, Cao Y, et al., 2024. Log Volume Measurement and Counting Based on Improved Cascade Mask R-CNN and Deep SORT. Forests, 15(11): 1884. https://doi.org/10.3390/f15111884 Yu, C. J., Sun, Y. K., Cao, Y., et al., 2024. Log Volume Measurement and Counting Based on Improved Cascade Mask R-CNN and Deep SORT. Forests, 15(11): 1892. Yun, L., Zhang, X. X., Zheng, Y. X., et al., 2023. Enhancing the Accuracy of Landslide Detection in UAV Images Using an Improved Mask R-CNN Model: A Case Study of Sanming, China. Sensors, 23(9): 4287. https://doi.org/ 10.3390/s23094287 Zhang, W. X., Witharana, C., Liljedahl, A. K., et al., 2018. Deep Convolutional Neural Networks for Automated Characterization of Arctic Ice-Wedge Polygons in Very High Spatial Resolution Aerial Imagery. Remote Sensing, 10(9): 1487. https://doi.org/10.3390/rs10091487 Zhang, X. W., Jin, J. J., Lan, Z. Z., et al., 2020. ICENET: A Semantic Segmentation Deep Network for River Ice by Fusing Positional and Channel-Wise Attentive Features. Remote Sensing, 12(2): 221. https://doi.org/10.3390/rs12020221 Zhang, X. Y., Zhou, S. H., Huang, J., et al., 2025. High-Order Spatial Feature Extraction Based Small Target Detection for UAV Aerial Photographs. Computer Engineering and Applications. (in Chinese with English abstract). Zhao, H., 2018. The Research of Glacial Lake Extraction Based on Landsat-8 OLI Imagery in High Mountain Region of Asian (Dissertation). University of Chinese Academy of Sciences (Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences), Beijing (in Chinese with English abstract). Zhao, H., 2023. The Research of Detection and Risk Evaluation of Glacial Lakes Based on Feature-Learning Methods across High Mountain Asia (Dissertation). University of Chinese Academy of Sciences (Xi'an Institute of Optics and Precision Mechanics, Chinese Academy of Sciences), Beijing (in Chinese with English abstract). Zou, Q., Zhou, B., Yang, T., et al., 2024. Spatio-Temporal Differentiation Characteristics of Glacial Lake Outburst in the Himalayas. Earth Science, 49(11): 4047-4062 (in Chinese with English abstract). 曹昱, 2022. 基于深度卷积神经网络的喜马拉雅冰湖自动检测方法研究(硕士学位论文). 成都: 电子科技大学. 车彦军, 陈丽花, 吴佳康, 等, 2024. 青藏高原冰前湖与冰川相互作用研究进展. 气候变化研究进展, 20(5): 519-533. 陈方, 王金晓, 张美美, 等, 2023. 基于历史边界的喜马拉雅山脉冰湖提取方法对比研究. 冰川冻土, 45(4): 1413-1427. 陈宫燕, 李婷, 陈军, 等, 2023. 基于栅格径流汇流模拟的西藏林芝市泥石流灾害预警模型初探. 中国地质灾害与防治学报, 34(1): 110-120. 陈浩文, 2022. 青藏高原冰湖溃决易发性评价及危险性预测研究(硕士学位论文). 成都: 成都理工大学. 成顺, 李建荣, 王志乾, 等, 2024. 基于YOLOv8轻量化水下光学图像识别算法. 激光与光电子学进展, 62(4): 402-412. 窦杰, 向子林, 许强, 等, 2023. 机器学习在滑坡智能防灾减灾中的应用与发展趋势. 地球科学, 48(5): 1657-1674. doi: 10.3799/dqkx.2022.419 郭慧琳, 谢元礼, 胡李发, 等, 2024. 改进HarDNet-MSEG的遥感影像水体信息提取方法. 地球信息科学学报, 26(7): 1745-1762. 胡家瑜, 2024. 藏东南典型冰碛湖演化及潜在溃决洪水模拟: 以炯拉错为例(硕士学位论文). 兰州: 西北师范大学. 蒋紫云, 2023. 气候变化背景下克亚吉尔冰川运动对冰湖变化的影响研究(硕士学位论文). 信阳: 信阳师范学院. 李均力, 盛永伟, 骆剑承, 2011. 喜马拉雅山地区冰湖信息的遥感自动化提取. 遥感学报, 15(1): 29-43. 李梦帆, 郑江华, 钱安良, 等, 2024. 基于决策树的天山冰湖提取方法研究. 干旱区研究, 41(10): 1699-1707. 李晓恩, 刘易, 江利明, 等, 2024. 山地冰川冰湖变化及其溃决洪水遥感监测研究进展. 地球信息科学学报, 26(4): 1019-1039. 李宇宸, 张军, 刘陈立, 2021. Sentinel-2影像的云南千湖山细小冰湖提取方法. 测绘科学, 46(4): 114-120. 刘美, 2020. Bhote Koshi流域冰湖溃决成灾机制与危险性评估(博士学位论文). 成都: 中国科学院大学(中国科学院水利部成都山地灾害与环境研究所). 刘帅琪, 李均力, 李若楠, 等, 2024. 基于Sentinel-2影像的新疆冰湖制图及空间分布特征. 冰川冻土, 46(2): 513-524. 鲁鸣, 2022. 标注受限情形下的高分辨率遥感图像水体提取方法研究(硕士学位论文). 长沙: 湖南大学. 马劲松, 2022. 基于多源遥感的全球冰湖时空变化监测(硕士学位论文). 南京: 南京信息工程大学. 苏步宇, 杜小平, 慕号伟, 等, 2024. 耦合Mask R-CNN和注意力机制的建筑物提取及后处理策略. 遥感技术与应用, 39(3): 620-632. 魏翠婷, 赵唯坚, 孙博超, 等, 2024. 基于改进Mask R-CNN与双目视觉的智能配筋检测. 浙江大学学报(工学版), 58(5): 1009-1019. 杨成德, 王欣, 魏俊峰, 等, 2019. 基于3S技术方法的中国冰湖编目. 地理学报, 74(3): 544-556. 杨泞滔, 聂勇, 2024. 一种改进的深度学习冰湖遥感制图方法及应用. 航天返回与遥感, 45(1): 41-52. 尹力辰, 王欣, 殷永胜, 等, 2024. 基于深度学习和Sentinel-2影像的冰湖自动提取. 遥感技术与应用, 39(6): 1319-1329. 张轩宇, 周思航, 黄健, 等, 2025. 基于高阶空间特征提取的无人机航拍小目标检测算法. 计算机工程与应用, 61(12): 210-221. 赵航, 2018. 基于Landsat-8遥感影像的高亚洲地区冰湖提取方法研究(硕士学位论文). 北京: 中国科学院大学(中国科学院遥感与数字地球研究所). 赵航, 2023. 基于特征学习的高亚洲地区冰湖识别及溃决风险研究(博士学位论文). 北京: 中国科学院大学(中国科学院西安光学精密机械研究所). 邹强, 周斌, 杨涛, 等, 2024. 喜马拉雅高海拔山区冰湖溃决时空分异特征. 地球科学, 49(11): 4047-4062. doi: 10.3799/dqkx.2024.083 -
点击查看大图
图(5) / 表(3)
计量
- 文章访问数: 1409
- HTML全文浏览量: 451
- PDF下载量: 71
- 被引次数: 0
