综合InSAR与深度学习的青海省化隆县活动滑坡自动识别

陈文雪; 王贤敏; 郭海湘; 曹里; 李冬冬; 隋兵

doi:10.11834/jrs.20255155

InSAR创新应用 | 浏览量 : 0 下载量: 39 CSCD: 0

R-PDF
PDF
导出
分享
收藏
专辑

综合InSAR与深度学习的青海省化隆县活动滑坡自动识别
Automatic identification of active landslides in Hualong County， Qinghai Province， integrating InSAR and deep learning
2026年30卷第4期页码：1198-1217
收稿：2025-05-13，

纸质出版：2026-04-07
DOI： 10.11834/jrs.20255155
稿件说明：

移动端阅览

陈文雪，王贤敏，郭海湘，曹里，李冬冬，隋兵.2026.综合InSAR与深度学习的青海省化隆县活动滑坡自动识别.遥感学报，30（4）： 1198-1217 DOI： 10.11834/jrs.20255155.

Chen W X，Wang X M，Guo H X，Cao L，Li D D and Sui B. 2026. Automatic identification of active landslides in Hualong County， Qinghai Province， integrating InSAR and deep learning. National Remote Sensing Bulletin， 30（4）：1198-1217 DOI： 10.11834/jrs.20255155.

摘要

灾难性滑坡往往由活动滑坡发展而来，活动滑坡的早期识别能够有效支撑灾害及时防治，避免人员伤亡与财产损失。干涉合成孔径雷达InSAR（Interferometric Synthetic Aperture Radar）技术为活动滑坡识别提供了重要支撑，然而目前活动滑坡识别主要依据地表形变、地形地貌特征，依靠目视解译，在广阔区域应用时面临耗时、高漏检、高虚警的瓶颈。本文综合InSAR与空间分析方法自动提取地表有效形变区域，构建了涵盖孕灾、致灾、形变特征的隐患识别综合判据，建立了充分挖掘孕灾、致灾与滑坡形变之间多尺度非线性关系的深度学习AMRetNet算法，在滑坡活跃的广阔化隆县区域开展活动滑坡自动识别，识别出活动滑坡178处，新发现活动滑坡48处，提出的AMRetNet算法性能优于目前活动滑坡自动识别领域典型算法，包括Transformer、U-Net、CART和SVM。本文工作为大范围城镇区域活动滑坡自动识别和早期发现提供了重要支撑。

Abstract

Large-scale landslide disasters often evolve from active landslides. Therefore

early and accurate identification of such active landslides is a key step to reducing disaster risks effectively. The core goal is to avoid casualties and minimize significant economic losses. At present

landslide disasters occur frequently in China. However

traditional manual interpretation methods have problems

such as low efficiency and insufficient identification accuracy. As a result

the actual needs of large-scale geological disaster monitoring cannot be adequately met. Accordingly

this study is dedicated to developing an automatic identification method suitable for large-scale active landslides. Moreover

Hualong County in Qinghai Province

which is severely affected by landslide disasters

is selected as a typical research area. Accurate and efficient identification of active landslides in the region is achieved by constructing a deep learning model. This approach ultimately provides a solid scientific basis for early warning and prevention of regional geological disasters and helps improve the overall ability of China’s mountainous cities and towns to prevent and control geological disaster risks.

This study comprehensively utilizes the small baseline subset interferometric synthetic aperture radar and spatial analysis methods to extract effective deformation areas on the ground surface automatically. It also constructs a comprehensive criterion for identifying hidden dangers related to active deformation

geology

terrain

environment

meteorology

and human engineering activities. This criterion covers landslide movement

disaster breeding

and disaster-causing characteristics. A deep learning AMRetNet algorithm that can fully explore the multiscale nonlinear relationship between disaster breeding

disaster-causing characteristics

and landslide deformation is established to capture the global context and local detail features effectively through the retention mechanism. An arithmetic module is also introduced to learn multiscale nonlinear relationships adaptively. The establishment and introduction of the aforementioned algorithm and module significantly improve the accuracy and robustness of landslide identification in complex geological environments.

In the case of Hualong County

this study successfully identified 178 active landslides

including 48 newly discovered ones. This result fully demonstrates the effectiveness of the method. The model exhibits excellent performance in various evaluation metrics in the test area (884 km²)

with accuracy

precision

F1-score

AUC

and kappa coefficients reaching 99.05%

90.21%

0.7332

0.9803

and 0.7286

respectively. It also has a low missed detection rate of 1.85%. Comparisons with mainstream algorithms

such as transformer

U-Net

CART

and SVM

as well as ablation experiments

showed that AMRetNet performed optimally in various performance metrics. This result demonstrates AMRetNet’s significant progressiveness and reliability.

This study confirms the superiority of the proposed automatic identification method for active landslides in accurately identifying active landslides in vast areas. The technical bottleneck of high false alarms and high missed detections in traditional methods is successfully solved by constructing a comprehensive identification index set and a deep learning AMRetNet algorithm that can fully explore the multiscale spatial nonlinear dependence between active landslide disaster breeding

disaster-causing characteristics

and deformation. The research results can provide reliable technical support for the prevention and control of geological disaster risks in mountainous towns in China. They also have important theoretical significance and promotion application value.

关键词

Keywords

references

Berardino P , Fornaro G , Lanari R and Sansosti E . 2002 . A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms . IEEE Transactions on Geoscience and Remote Sensing , 40 ( 11 ): 2375 - 2383 [ DOI: 10.1109/TGRS.2002.803792 http://dx.doi.org/10.1109/TGRS.2002.803792 ]

Cai J H , Zhang L , Dong J , Guo J C , Wang Y A and Liao M S . 2023 . Automatic identification of active landslides over wide areas from time-series InSAR measurements using Faster RCNN . International Journal of Applied Earth Observation and Geoinformation , 124 : 103516 [ DOI: 10.1016/j.jag.2023.103516 http://dx.doi.org/10.1016/j.jag.2023.103516 ]

Chen X M , Yao X , Zhou Z K , Liu Y , Yao C C and Ren K Y . 2022 . DRs-UNet: a deep semantic segmentation network for the recognition of active landslides from InSAR imagery in the three rivers region of the Qinghai-Tibet Plateau . Remote Sensing , 14 ( 8 ): 1848 [ DOI: 10.3390/rs14081848 http://dx.doi.org/10.3390/rs14081848 ]

Chen X R , Zhao C Y , Liu X J , Zhang S C , Xi J B and Khan B A . 2024 . An embedding Swin transformer model for automatic slow-moving landslide detection based on InSAR products . IEEE Transactions on Geoscience and Remote Sensing , 62 : 5223915 [ DOI: 10.1109/TGRS.2024.3470325 http://dx.doi.org/10.1109/TGRS.2024.3470325 ]

Cheng Y , Hu R J , Ying H C , Shi X , Wu J and Lin W . 2024 . Arithmetic feature interaction is necessary for deep tabular learning // Proceedings of the 38th AAAI Conference on Artificial Intelligence . Vancouver : AAAI: 11516 - 11524 [ DOI: 10.1609/aaai.v38i10.29033 http://dx.doi.org/10.1609/aaai.v38i10.29033 ]

Ciuffi P , Bayer B , Berti M , Franceschini S and Simoni A . 2024 . InSAR stacking to detect active landslides and investigate their relation to rainfalls in the Northern Apennines of Italy . Geomorphology , 457 : 109242 [ DOI: 10.1016/j.geomorph.2024.109242 http://dx.doi.org/10.1016/j.geomorph.2024.109242 ]

Hendrycks D and Gimpel K . 2023 . Gaussian error linear units (GELUs) . arXiv preprint arXiv : 1606 . 08415 [ DOI: 10.48550/arXiv.1606.08415 http://dx.doi.org/10.48550/arXiv.1606.08415 ]

Hu Y R , Cao R T , Qi Y H and Li B . 2012 . On sliding slope mechanism and risk zoning at Hualong Area in Qinghai . Shanxi Architecture , 38 ( 27 ): 88 - 89

胡云瑞 , 曹荣泰 , 祈永辉 , 李彬 . 2012 . 青海省化隆地区滑坡机制及危险性分区 . 山西建筑 , 38 ( 27 ): 88 - 89 [ DOI: 10.13719/j.cnki.cn14-1279/tu.2012.27.006 http://dx.doi.org/10.13719/j.cnki.cn14-1279/tu.2012.27.006 ]

Lan H N . 2021 . Intelligent Identification of Geological Landslide based on Multi-Source Data Fusion . Beijing : Beijing University of Posts and Telecommunications

冉号楠 . 2021 . 基于多源数据融合的地质滑坡智能判识研究 . 北京 : 北京邮电大学 [ DOI: 10.26969/d.cnki.gbydu.2021.000271 http://dx.doi.org/10.26969/d.cnki.gbydu.2021.000271 ]

Li X Q , Li W L , Wu Z L , Xu Q , Zheng D , Dong X J , Lu H Y , Shan Y F , Zhou S S , Yu W L and Wang X C . 2024 . Identification and deformation characteristics of active landslides at large hydropower stations at the early impoundment stage: a case study of the Lianghekou reservoir area in Sichuan Province, Southwest China . Remote Sensing , 16 ( 17 ): 3175 [ DOI: 10.3390/rs16173175 http://dx.doi.org/10.3390/rs16173175 ]

Liang Y W , Zhang Y , Li Y X and Xiong J Q . 2023 . Automatic identification for the boundaries of InSAR anomalous deformation areas based on semantic segmentation model . Remote Sensing , 15 ( 21 ): 5262 [ DOI: 10.3390/rs15215262 http://dx.doi.org/10.3390/rs15215262 ]

Liu X J , Zhao C Y , Li B , Wang W D , Zhang Q , Gao Y , Chen L Q , Wang B H , Hao J M and Yang X H . 2025 . Identification and dynamic deformation monitoring of active landslides in Jishishan earthquake area (Gansu, China) using InSAR technology . Geomatics and Information Science of Wuhan University , 50 ( 2 ): 297 - 312

刘晓杰 , 赵超英 , 李滨 , 王文达 , 张勤 , 高杨 , 陈立权 , 王宝行 , 郝君明 , 杨校辉 . 2025 . 基于InSAR技术的甘肃积石山震区活动滑坡识别与动态形变监测 . 武汉大学学报(信息科学版) , 50 ( 2 ): 297 - 312 [ DOI: 10.13203/j.whugis20240054 http://dx.doi.org/10.13203/j.whugis20240054 ]

Liu X L , Zhang X J , Li J L , Wang Y F , Zhang X G , Yuan X N , Wang K Y , Wang C G , Liu J and Hou E G . 2023 . Characteristics and scientific values of “Canyon and Danxia” landform in Hualong county, Qinghai Province . Geoscience , 37 ( 1 ): 233 - 244

刘心兰 , 张绪教 , 李俊磊 , 王一凡 , 张向格 , 袁晓宁 , 王凯雅 , 王重歌 , 刘江 , 侯恩刚 . 2023 . 青海化隆县独特的峡谷丹山地貌景观及其科学价值 . 现代地质 , 37 ( 1 ): 233 - 244 [ DOI: 10.19657/j.geoscience.1000-8527.2023.031 http://dx.doi.org/10.19657/j.geoscience.1000-8527.2023.031 ]

Liu Y , Yao X , Gu Z K , Li R J , Zhou Z K , Liu X H , Jiang S , Yao C C and Wei S F . 2024 . Research on automatic recognition of active landslides using InSAR deformation under digital morphology: a case study of the Baihetan reservoir, China . Remote Sensing of Environment , 304 : 114029 [ DOI: 10.1016/j.rse.2024.114029 http://dx.doi.org/10.1016/j.rse.2024.114029 ]

Nefros C , Alatza S , Loupasakis C and Kontoes C . 2023 . Persistent Scatterer Interferometry (PSI) technique for the identification and monitoring of critical landslide areas in a regional and mountainous road network . Remote Sensing , 15 ( 6 ): 1550 [ DOI: 10.3390/rs15061550 http://dx.doi.org/10.3390/rs15061550 ]

Ramachandran P , Zoph B and Le Q V . 2017 . Swish: a self-gated activation function . arXiv preprint arXiv : 1710 . 05941 [ DOI: 10.48550/arXiv.1710.05941 http://dx.doi.org/10.48550/arXiv.1710.05941 ]

Rosenblatt M . 1956 . Remarks on some nonparametric estimates of a density function . The Annals of Mathematical Statistics , 27 ( 3 ): 832 - 837 [ DOI: 10.1214/aoms/1177728190 http://dx.doi.org/10.1214/aoms/1177728190 ]

Samsonov S and Blais-Stevens A . 2023 . Satellite interferometry for regional assessment of landslide hazard to pipelines in northeastern British Columbia, Canada . International Journal of Applied Earth Observation and Geoinformation , 118 : 103273 [ DOI: 10.1016/j.jag.2023.103273 http://dx.doi.org/10.1016/j.jag.2023.103273 ]

Silverman B W . 1986 . Density Estimation for Statistics and Data Analysis . New York : Chapman and Hall

Sun Y T , Dong L , Huang S H , Ma S M , Xia Y Q , Xue J L , Wang J Y and Wei F R . 2023 . Retentive network: a successor to transformer for large language models . arXiv preprint arXiv : 2307 . 08621 [ DOI: 10.48550/arXiv.2307.08621 http://dx.doi.org/10.48550/arXiv.2307.08621 ]

Tang H B , Chen Q , He Z L and Li D J . 2019 . Evaluation and analysis of groundwater resources in small watersheds of Hualong county of Qinghai Province . Yellow River , 41 ( 9 ): 70 - 72 , 91

唐洪波 , 陈强 , 何灼伦 , 李德靖 . 2019 . 青海省化隆县小流域地下水资源量评价 . 人民黄河 , 41 ( 9 ): 70- 72 , 91 [ DOI: 10.3969/j.issn.1000-1379.2019.09.014 http://dx.doi.org/10.3969/j.issn.1000-1379.2019.09.014 ]

Vaswani A , Shazeer N , Parmar N , Uszkoreit J , Jones L , Gomez A N , Kaiser Ł and Polosukhin I . 2017 . Attention is all you need // Proceedings of the 31st International Conference on Neural Information Processing Systems . Long Beach : Curran Associates Inc.: 6000 - 6010

Wang M , Wei J , Wang X , et al . Reconstruction of all-sky daily air temperature datasets with high accuracy in China from 2003 to 2022 [J ] . Scientific data , 2024 , 11 ( 1 ): 1133 .

Wei C H , Li W L , Wu Z L , Zhou S S , Wang X C , Li X Q and Li Y S . 2025 . Identification of active landslides and analysis of deformation influencing factors in the Baihetan Reservoir area . Bulletin of Geological Science and Technology , 44 ( 4 ): 62 - 77

韦春豪 , 李为乐 , 吴章雷 , 周胜森 , 王新程 , 李雪清 , 李雨森 . 2025 . 白鹤滩库区活动滑坡识别及形变影响因素分析 . 地质科技通报 , 44 ( 4 ): 62 - 77 [ DOI: 10.19509/j.cnki.dzkq.tb20240351 http://dx.doi.org/10.19509/j.cnki.dzkq.tb20240351 ]

Wu Q , Ge D Q , Yu J C , Zhang L , Li M , Liu B , Wang Y , Ma Y N and Liu H J . 2022 . Deep learning identification technology of InSAR significant deformation zone of potential landslide hazard at large scale . Acta Geodaetica et Cartographica Sinica , 51 ( 10 ): 2046 - 2055

吴琼 , 葛大庆 , 于峻川 , 张玲 , 李曼 , 刘斌 , 王艳 , 马燕妮 , 刘宏娟 . 2022 . 广域滑坡灾害隐患InSAR显著性形变区深度学习识别技术 . 测绘学报 , 51 ( 10 ): 2046 - 2055 [ DOI: 10.11947/j.AGCS.2022.20220303 http://dx.doi.org/10.11947/j.AGCS.2022.20220303 ]

Xu L , Ju N P , Deng M D , Xie M L , Zhou F J , Li C G and Tang Y . 2024 . Comparative analysis of landslide susceptibility assessment accuracy in different evaluation units . Journal of Engineering Geology , 32 ( 5 ): 1640 - 1653

许烈 , 巨能攀 , 邓明东 , 解明礼 , 周福军 , 李朝纲 , 唐颍 . 2024 . 不同评价单元的滑坡易发性评价精度对比分析 . 工程地质学报 , 32 ( 5 ): 1640 - 1653 [ DOI: 10.13544/j.cnki.jeg.2024-0341 http://dx.doi.org/10.13544/j.cnki.jeg.2024-0341 ]

Xu Q , Lu H Y , Li W L , Dong X J and Guo C . 2022 . Types of Potential Landslide and Corresponding Identification Technologies . Geomatics and Information Science of Wuhan University , 47 ( 3 ): 377 - 387

许强 , 陆会燕 , 李为乐 , 董秀军 , 郭晨 . 2022 . 滑坡隐患类型与对应识别方法 . 武汉大学学报(信息科学版) , 47 ( 3 ): 377 - 387 [ DOI: 10.13203/j.whugis20210618 http://dx.doi.org/10.13203/j.whugis20210618 ]

Xun Z Y , Zhao C Y , Kang Y , Liu X J , Liu Y Y and Du C Y . 2022 . Automatic extraction of potential landslides by integrating an optical remote sensing image with an InSAR-derived deformation map . Remote Sensing , 14 ( 11 ): 2669 [ DOI: 10.3390/rs14112669 http://dx.doi.org/10.3390/rs14112669 ]

Yan G , Liang S Y and Zhao H L . 2017 . An approach to improving slope unit division using GIS technique . Scientia Geographica Sinica , 37 ( 11 ): 1764 - 1770

颜阁 , 梁收运 , 赵红亮 . 2017 . 基于GIS的斜坡单元划分方法改进与实现 . 地理科学 , 37 ( 11 ): 1764 - 1770 [ DOI: 10.13249/j.cnki.sgs.2017.11.019 http://dx.doi.org/10.13249/j.cnki.sgs.2017.11.019 ]

Yao X , Deng J H , Liu X H , Zhou Z K , Yao J M , Dai F C , Ren K Y and Li L J . 2020 . Primary recognition of active landslides and development rule analysis for pan Three-river-parallel Territory of Tibet Plateau . Advanced Engineering Sciences , 52 ( 5 ): 16 - 37

姚鑫 , 邓建辉 , 刘星洪 , 周振凯 , 姚佳明 , 戴福初 , 任开瑀 , 李凌婧 . 2020 . 青藏高原泛三江并流区活动性滑坡InSAR初步识别与发育规律分析 . 工程科学与技术 , 52 ( 5 ): 16 - 37 [ DOI: 10.15961/j.jsuese.202000529 http://dx.doi.org/10.15961/j.jsuese.202000529 ]

Zhang C L , Li Z H , Yu C , Chen B , Ding M T , Zhu W , Yang J , Liu Z J and Peng J B . 2022a . An integrated framework for wide-area active landslide detection with InSAR observations and SAR pixel offsets . Landslides , 19 ( 12 ): 2905 - 2923 [ DOI: 10.1007/s10346-022-01954 http://dx.doi.org/10.1007/s10346-022-01954 ]

Zhang K X . 2005 . Regional geological survey report of Minhe Hui and Tu Autonomous County map sheet J 48 C 004001 ( 1 : 250, 000 Scale): volumes I & II. China University of Geosciences (Wuhan)

张克信 . 2005 . 民和回族土族自治县幅J 48 C 004001 1 / 25 万区域地质调查报告: 上、下册. 中国地质大学(武汉) [ DOI: 10.35080/n01.c.122809 http://dx.doi.org/10.35080/n01.c.122809 ]

Zhang K X . 2006 . Re-survey regional geological report of Linxia city map sheet I 48 C 001001 (1: 250, 000 Scale). China University of Geosciences (张克信 . 2006. 临夏市幅I 48 C 001001 1 / 25 万区域地质调查报告: 修测. 中国地质大学) [ DOI: 10.35080/n01.c.123677 http://dx.doi.org/10.35080/n01.c.123677 ]

Zhang T , Zhang W C , Cao D , Yi Y N and Wu X . 2022b . A new deep learning neural network model for the identification of InSAR anomalous deformation areas . Remote Sensing , 14 ( 11 ): 2690 [ DOI: 10.3390/rs14112690 http://dx.doi.org/10.3390/rs14112690 ]