模拟困难样本的Mask R-CNN滑坡分割识别

姜万冬; 席江波; 李振洪; 丁明涛; 杨立功; 谢大帅

doi:10.13203/j.whugis20200692

模拟困难样本的Mask R-CNN滑坡分割识别

doi: 10.13203/j.whugis20200692

姜万冬^1,2,
席江波^1,2,3,
李振洪^1,2,3,
丁明涛^1,2,3,
杨立功¹,
谢大帅^1,2

1 长安大学地质工程与测绘学院, 陕西西安, 710054;
2 长安大学地学与卫星大数据研究中心, 陕西西安, 710054;
3 西部矿产资源与地质工程教育部重点实验室, 陕西西安, 710054

基金项目:

国家自然科学基金重大项目（41941019），国家重点研发计划（2018YFC1504805），国家自然科学基金（61806022，41874005），中央高校基本科研业务费专项资金（300102260301/087，300102260404/087，300102269103，300102269304和300102269205），地理信息国家重点实验室开放基金（SKLGIE2018-M-3-4）。

详细信息

作者简介:
姜万冬,硕士,主要研究方向为遥感影像地质灾害智能解译,CHD_jwd@126.com*

Landslide Detection and Segmentation Using Mask R-CNN with Simulated Hard Samples

1 College of Geological Engineering and Geomantics, Chang'an University, Xi'an, Shaanxi 710054, China;
2 Big Data Center for Geosciences and Satellites, Chang'an University, Xi'an, Shaanxi 710054, China;
3 Key Laboratory of Western China's Mineral Resources and Geological Engineering, Ministry of Education, Xi'an, Shaanxi 710054, China

Funds:

Major Program of the National Natural Science Foundation of China (41941019), The National Key Research and Development Program of China (2018YFC1504805), The National Natural Science Foundation of China(61806022, 41874005), The Fundamental Research Funds for the Central Universities (300102260301/087, 300102260404/087, 300102269103, 300102269304, and 300102269205), the State Key Laboratory of Geographic Information (SKLGIE2018-M-3-4).

摘要: 随着人工智能的发展，利用高分影像进行滑坡等地质灾害识别逐渐成为研究热点。滑坡目视解译需依赖专家经验，传统滑坡自动识别方法又易将滑坡和裸地、道路等地物混淆。针对以上问题，提出基于模拟困难样本的Mask R-CNN（mask region- based convolutionalnetwork）滑坡提取方法。在现有样本的基础上，利用滑坡的形状、颜色、纹理等特征模拟更为复杂的滑坡背景，进行困难样本挖掘增强，并将得到的困难样本输入Mask R-CNN网络进行滑坡精细检测分割。在实际研究区域中，由于滑坡数量有限，因此在频率域进行小样本学习，在减少数据需求的同时，保证分割识别的准确度。在贵州毕节的实验结果表明，基于模拟困难样本的Mask R-CNN方法检测精度为94.0%，像素分割平均准确率为90.3%，可实现低虚警率下的高性能检测分割；采用频率域学习，在一半数据输入量的情况下，模型检测精度仍可得到提升；并利用天水地区的滑坡区域进行实际验证，进一步证明了所提方法的有效性。
- 深度学习 /
- 滑坡提取 /
- Mask R-CNN /
- 困难样本 /
- 频率域
Abstract: With the advance in artificial intelligence, using high-resolution images to detect geological hazards has gradually become a research hotspot. Visual interpretation of landslides heavily relies on expert experience, and conventional automatic landslide detection approaches are sensitive to the presence of bare land, roads and other ground objects. To address these, a Mask R-CNN with simulated hard samples is presented in this paper for landslide detection and segmentation. Based on existing landslide samples, hard landslide samples are simulated by utilizing the shapes, colors, textures, and other characteristics of landslides to make each of the samples with a more complicated background. The original imagery and simulated hard samples are then fed into the Mask R-CNN for landslide detection and segmentation. Since the number of landslides is often limited in reality, small sample learning in the frequency domain is also presented in this paper to reduce the number of input samples while ensuring the accuracy of detection and segmentation. The experimental results in Bijie, Guizhou Province, showed that the detection and the average pixel segmentation accuracies of the proposed Mask R-CNN method with simulated hard samples are 94.0% and 90.3%, respectively. It is seen that the proposed method has high performance on landslide detection and segmentation with low false alarm rates. In addition, the performance of the proposed small-sample-based learning method in frequency domain can be improved even with a half of the data input. The effectiveness of the proposed Mask R-CNN method is further proved by the successful detection of Tianshui landslides in Gansu Province.
- Deep learning /
- Landslide detection /
- Mask R-CNN /
- Hard samples /
- Frequency domain

[1]	Dai K, Li Z, Xu Q, et al. Entering the Era of Earth Observation-Based Landslide Warning Systems:A Novel and Exciting Framework[J]. IEEE Geoscience and Remote Sensing Magazine, 2020, 8(1):136-153.
[2]	BAI Zhengwei Z Q, HUANG Guanwen, JING Ce, WANG Jiaxing. Real-time BeiDou landslide monitoring technology of-light terminal plus industry cloud‖[J]. Acta Geodaetica et Cartographica Sinica, 2019, 48(11):1424-1429.
[3]	Bovenga F, Nitti D, Fornaro G, et al. Using C/X-band SAR interferometry and GNSS measurements for the Assisi landslide analysis[J]. International Journal of Remote Sensing, 2013, 34(11):4083-4104.
[4]	Qiu D, Wang L, Luo D, et al. Landslide monitoring analysis of single-frequency BDS/GPS combined positioning with constraints on deformation characteristics[J]. Survey Review, 2019, 51(367):364-372.
[5]	Jaboyedoff M, Oppikofer T, Abellan A, et al. Use of LIDAR in landslide investigations:a review[J]. Natural Hazards, 2012, 61(1):5-28.
[6]	Mckean J, Roering J J. Objective landslide detection and surface morphology mapping using high-resolution airborne laser altimetry[J]. Geomorphology, 2004, 57(3):331-351.
[7]	Sato H P, Hasegawa H, Fujiwara S, et al. Interpretation of landslide distribution triggered by the 2005 Northern Pakistan earthquake using SPOT 5 imagery[J]. Landslides, 2007, 4(2):113-122.
[8]	Duric U, Marjanovic M, Radic Z, et al. Machine learning based landslide assessment of the Belgrade metropolitan area:Pixel resolution effects and a cross-scaling concept[J]. Engineering Geology, 2019, 256:23-38.
[9]	Lu P, Qin Y, Li Z, et al. Landslide mapping from multi-sensor data through improved change detection-based Markov random field[J]. Remote Sensing of Environment, 2019, 231:111235.
[10]	Pawluszek K, Marczak S, Borkowski A, et al. Multi-Aspect Analysis of Object-Oriented Landslide Detection Based on an Extended Set of LiDAR-Derived Terrain Features[J]. Isprs International Journal of Geo-Information, 2019, 8(8):321.
[11]	Bacha, Werff V D, Shafique, et al. Transferability of object-based image analysis approaches for landslide detection in the Himalaya Mountains of northern Pakistan[J]. International Journal of Remote Sensing, 2020, 41(9):3390-3410.
[12]	Petschko H, Bell R, Glade T. Effectiveness of visually analyzing LiDAR DTM derivatives for earth and debris slide inventory mapping for statistical susceptibility modeling[J]. Landslides, 2016, 13(5):857-872.
[13]	Danneels G, Pirard E, Havenith H B. Automatic landslide detection from remote sensing images using supervised classification methods[C], IEEE International Geoscience & Remote Sensing Symposium, Barcelona, Spain, 2007.
[14]	Jones J W, Desmond G B, Glover C H, et al. An approach to regional wetland digital elevation model development using a differential global positioning system and a custom-built helicopter-based surveying system[J]. International Journal of Remote Sensing, 2012, 33(2):450-465.
[15]	Cao Y, Yin K, Zhou C, et al. Establishment of Landslide Groundwater Level Prediction Model Based on GA-SVM and Influencing Factor Analysis[J]. Sensors (Basel, Switzerland), 2020, 20(3):845.
[16]	Zhu C H, Hu G D. Time Series Prediction of Landslide Displacement Using SVM Model:Application to Baishuihe Landslide in Three Gorges Reservoir Area, China[J]. Applied Mechanics & Materials, 2013, 239-240:1413-1420.
[17]	Zhang K, Wu X, Niu R, et al. The assessment of landslide susceptibility mapping using random forest and decision tree methods in the Three Gorges Reservoir area, China[J]. Environmental Earth Sciences, 2017, 76(10):405.
[18]	Stumpf A, Kerle N. Object-oriented mapping of landslides using Random Forests[J]. Remote Sensing of Environment, 2011, 115(10):2564-2577.
[19]	Mezaal M R, Pradhan B, Rizeei H M. Improving Landslide Detection from Airborne Laser Scanning Data Using Optimized Dempster-Shafer[J]. Remote Sensing, 2018, 10(7):1029.
[20]	Wang F, Jiang M, Qian C, et al. Residual Attention Network for Image Classification[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 2017.
[21]	Zhao Z, Zheng P, Xu S, et al. Object Detection With Deep Learning:A Review[J]. IEEE Transactions on Neural Networks and Learning Systems, 2019, 30(11):3212-3232.
[22]	Ronneberger O, Fischer P, Brox T. U-Net:Convolutional Networks for Biomedical Image Segmentation[M]//NAVAB N, HORNEGGER J, WELLS W M, et al. Medical Image Computing and Computer-Assisted Intervention, Pt Iii. 2015:234-241.
[23]	Shi W, Zhang M, Ke H, et al. Landslide Recognition by Deep Convolutional Neural Network and Change Detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 1-19.
[24]	Ghorbanzadeh O, Blaschke T, Gholamnia K, et al. Evaluation of Different Machine Learning Methods and Deep-Learning Convolutional Neural Networks for Landslide Detection[J]. Remote Sensing, 2019, 11(2):196.
[25]	Yu B, Chen F, Xu C. Landslide detection based on contour-based deep learning framework in case of national scale of Nepal in 2015[J]. Computers & Geoences, 2019, 135:104388.
[26]	Wang H, Zhang L, Yin K, et al. Landslide identification using machine learning[J]. Geoence Frontiers, 2020, 08(07):9705-9726.
[27]	Liu P, Wei Y, Wang Q, et al. Research on Post-Earthquake Landslide Extraction Algorithm Based on Improved U-Net Model[J]. Remote Sensing, 2020, 12(5):894.
[28]	Fukushima K, Miyake S, Ito T. Neocognitron:A neural network model for a mechanism of visual pattern recognition[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1983, SMC-13(5):826-834.
[29]	Girshick R, Donahue J, Darrell T, et al. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation[C], 2014 IEEE Conference on Computer Vision and Pattern Recognition, OH, USA, 2014.
[30]	Girshick R. Fast R-CNN[C], 2015 IEEE International Conference on Computer Vision (ICCV), Boston, MA, 2015.
[31]	Ren S, He K, Girshick R, et al. Faster R-CNN:Towards Real-Time Object Detection with Region Proposal Networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6):1137-1149.
[32]	He K, Gkioxari G, Dollár P, et al. Mask R-CNN[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2):386-397.
[33]	Bochkovskiy A, Wang C Y, Liao H Y M. YOLOv4:Optimal Speed and Accuracy of Object Detection[C]. Computer Vision and Pattern Recognition, Seattle, WA, USA, 2020.
[34]	Ji S, Yu D, Shen C, et al. Landslide detection from an open satellite imagery and digital elevation model dataset using attention boosted convolutional neural networks[J]. Landslides, 2020, 17:1337-1352.

[1]	郭从洲, 李可, 李贺, 童晓冲, 王习文. 遥感图像质量等级分类的深度卷积神经网络方法 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20200292
[2]	赵之若, 王少宇, 王心宇, 钟燕飞. 一种改进的火星车多光谱影像深度新颖目标探测方法 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20220119
[3]	李星华, 白学辰, 李正军, 左芝勇. 面向高分影像建筑物提取的多层次特征融合网络 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20210506
[4]	李彦胜, 张永军. 耦合知识图谱和深度学习的新一代遥感影像解译范式 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20210652
[5]	林学楷, 许才军. 深度学习驱动的地震目录构建：PhaseNet和EqT模型的对比与评估 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20220197
[6]	王艳东, 邵鑫, 刘波, 邓跃进, 魏广泽, 豆明宣. 一种利用Mask R-CNN的遥感影像与矢量数据配准方法 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20200137
[7]	张宇, 江鹏, 郭文飞, 张丹, 韩震. 一种利用两阶段学习模型的水下阵列定位方法 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20210466
[8]	范雕, 李姗姗, 孟书宇, 邢志斌, 张驰, 冯进凯, 曲政豪. 海底地形高次项对海面重力信息影响分析 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20190192
[9]	季顺平, 罗冲, 刘瑾. 基于深度学习的立体影像密集匹配方法综述 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20200620
[10]	陆川伟, 孙群, 赵云鹏, 孙士杰, 马京振, 程绵绵, 李元復. 一种基于条件生成式对抗网络的道路提取方法 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20190159
[11]	郭旦怀, 张鸣珂, 贾楠, 王彦棡. 融合深度学习技术的用户兴趣点推荐研究综述 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20200334
[12]	眭海刚, 黄立洪, 刘超贤. 利用具有注意力的Mask R-CNN检测震害建筑物立面损毁 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20200158
[13]	巨袁臻, 许强, 金时超, 李为乐, 董秀军, 郭庆华. 使用深度学习方法实现黄土滑坡自动识别 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20200132
[14]	潘银, 邵振峰, 程涛, 贺蔚. 利用深度学习模型进行城市内涝影响分析 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20170217
[15]	邵振峰, 张源, 黄昕, 朱秀丽, 吴亮, 万波. 基于多源高分辨率遥感影像的2 m不透水面一张图提取 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20180196
[16]	张兵. 遥感大数据时代与智能信息提取 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20180172
[17]	樊恒, 徐俊, 邓勇, 向金海. 基于深度学习的人体行为识别 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20140110
[18]	王智均, 李德仁, 李清泉. 利用小波变换对影像进行融合的研究 . 武汉大学学报 ● 信息科学版,
[19]	赫晓慧, 陈明扬, 李盼乐, 田智慧, 周广胜. 结合DCNN与短距条件随机场的遥感影像道路提取 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20210464
[20]	肖力炀, 李伟, 袁博, 崔逸群, 高荣, 王文庆. 一种改进型Mask R-CNN模型的路面裂缝识别方法 . 武汉大学学报 ● 信息科学版, doi: 10.13203/j.whugis20210279

点击查看大图

计量

文章访问数: 609
HTML全文浏览量: 163
PDF下载量: 73
被引次数: 0

姓名
邮箱
手机号码
标题
留言内容
验证码

留言板

模拟困难样本的Mask R-CNN滑坡分割识别

doi: 10.13203/j.whugis20200692

作者简介:
姜万冬,硕士,主要研究方向为遥感影像地质灾害智能解译,CHD_jwd@126.com*

Landslide Detection and Segmentation Using Mask R-CNN with Simulated Hard Samples

计量

模拟困难样本的Mask R-CNN滑坡分割识别

doi: 10.13203/j.whugis20200692

1 长安大学地质工程与测绘学院, 陕西西安, 710054;

2 长安大学地学与卫星大数据研究中心, 陕西西安, 710054;

3 西部矿产资源与地质工程教育部重点实验室, 陕西西安, 710054

作者简介:
姜万冬,硕士,主要研究方向为遥感影像地质灾害智能解译,CHD_jwd@126.com*

English Abstract

Landslide Detection and Segmentation Using Mask R-CNN with Simulated Hard Samples

全文HTML

目录

留言板

模拟困难样本的Mask R-CNN滑坡分割识别

doi: 10.13203/j.whugis20200692

作者简介: 姜万冬,硕士,主要研究方向为遥感影像地质灾害智能解译,CHD_jwd@126.com*

Landslide Detection and Segmentation Using Mask R-CNN with Simulated Hard Samples

计量

出版历程

模拟困难样本的Mask R-CNN滑坡分割识别

doi: 10.13203/j.whugis20200692

1 长安大学地质工程与测绘学院, 陕西 西安, 710054; 2 长安大学地学与卫星大数据研究中心, 陕西 西安, 710054; 3 西部矿产资源与地质工程教育部重点实验室, 陕西 西安, 710054

作者简介: 姜万冬,硕士,主要研究方向为遥感影像地质灾害智能解译,CHD_jwd@126.com*

English Abstract

Landslide Detection and Segmentation Using Mask R-CNN with Simulated Hard Samples

全文HTML

目录

作者简介:
姜万冬,硕士,主要研究方向为遥感影像地质灾害智能解译,CHD_jwd@126.com*

1 长安大学地质工程与测绘学院, 陕西西安, 710054;

2 长安大学地学与卫星大数据研究中心, 陕西西安, 710054;

3 西部矿产资源与地质工程教育部重点实验室, 陕西西安, 710054

作者简介:
姜万冬,硕士,主要研究方向为遥感影像地质灾害智能解译,CHD_jwd@126.com*