All Issue

2022 Vol.56, Issue 4 Preview Page

Research Article

Flood Detection from Sentinel-1 SAR Images using U-Net and HRNetV2 Models

U-Net과 HRNetV2 모델을 이용한 Sentinel-1 SAR 영상의 홍수탐지

Soyeon Choia
,
Ganghyun Parkb
,
Jonggu Kangc
,
Geunah Kimd
,
Youjeong Youne
,
Yang-Won Leef†
최 소연a
,
박 강현b
,
강 종구c
,
김 근아d
,
윤 유정e
,
이 양원f†
aMaster Student, Department of Spatial Information Engineering, Pukyong National University
bMaster Student, Department of Spatial Information Engineering, Pukyong National University
cMaster Student, Department of Spatial Information Engineering, Pukyong National University
dMaster Student, Department of Spatial Information Engineering, Pukyong National University
ePhD Student, Department of Spatial Information Engineering, Pukyong National University
fProfessor, Department of Spatial Information Engineering, Pukyong National University
a부경대학교 지구환경시스템과학부 공간정보공학전공 석사과정
b부경대학교 지구환경시스템과학부 공간정보공학전공 석사과정
c부경대학교 지구환경시스템과학부 공간정보공학전공 석사과정
d부경대학교 지구환경시스템과학부 공간정보공학전공 석사과정
e부경대학교 지구환경시스템과학부 공간정보공학전공 박사과정
f부경대학교 지구환경시스템과학부 공간정보공학전공 교수
*Corresponding Author
31 December 2022. pp. 409-419
PDF XML Citation
Abstract

Satellite remote sensing is a valuable tool for monitoring floods in a wide area. In particular, Synthetic Aperture Radar (SAR) has the advantage of the observation capability of penetrating clouds and can perform flood detection under rainy conditions. We conducted a flood detection from Sentinel-1 images using a large database (Flood Extent Detection) provided by the National Aeronautics and Space Administration (NASA) and deep learning models. A performance test using label and prediction images showed a precision of 0.956, a recall of 0.953, and a Mean Intersection over Union (mIOU) of 0.933. As future work, algorithm developments should be necessary for small flood areas, using Digital Elevation Model (DEM), State-of-the-Art (SOTA) models, and (Super Resolution, SR) techniques.

Keywords
Deep learning
Flood detection
SAR

위성원격탐사는 광역적인 홍수 감시의 수단으로 활용되고 있는데, 특히 합성개구레이더(Synthetic Aperture Radar, SAR)는 광학센서와 달리 구름을 투과하여 관측하기 때문에 우천 시에도 홍수 탐지가 가능한 장점이 있다. 본 연구에서는 미국 항공우주국(National Aeronautics and Space Administration, NASA) Flood Extent Detection 대용량 학습자료와 딥러닝 모델을 사용하여 Sentinel-1 영상의 홍수탐지를 수행하였다. 예측 영상과 레이블 영상을 비교하여 혼동행렬을 작성하고 성능평가를 수행한 결과, 정밀도 0.956, 재현율 0.953, 평균 교집합 대 합집합 비율(Intersection over Union, mIOU) 0.933의 상당히 높은 정확도 통계량을 나타냈다. 다만, 규모가 작은 범람 영역에 대한 정확도 향상이 필요한데, 이는 향후 연구에서 Digital Elevation Model (DEM) 등의 보조자료 사용과 최신(State-of-the-Art, SOTA) 모델의 활용 및 초해상화(Super Resolution, SR)의 적용 등을 통해 개선될 수 있을 것이다.

키워드
딥러닝
홍수탐지
합성개구레이더
References
1
Gahlot, S., I. Gurung, A. Molthan, M. Maskey, and M. Ramasubramanian, 2021, Flood Extent Data for Machine Learning Version 1.0, https://doi.org/10.34911/rdnt.ebk43 10.34911/rdnt.ebk43x
2
Global Change Research Program, 2018, Fourth National Climate Assessment: Chapter 2 Our Changing Climate, https://nca2018.globalchange.gov/chapter/2/
3
Huang, G., Z. Liu, L. van Der Maaten, and K. Weinberger, 2017, Densely connected convolutional networks, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 4700-4708. 10.1109/CVPR.2017.243PMC5598342
4
Jeon, H., D. Kim, and J. Kim, 2021, Water body detection using deep learning with Sentinel-1 SAR satellite data and land cover maps, 2021 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), 8495-8498. 10.1109/IGARSS47720.2021.9553555
5
Li, Y., S. Martinis, and M. Wieland, 2019, Urban flood mapping with an active self-learning convolutional neural network based on TerraSAR-X intensity and interferometric coherence, ISPRS Journal of Photogrammetry and Remote Sensing, 152: 178-191. 10.1016/j.isprsjprs.2019.04.014
6
Liang, J., and D. Liu, 2020, A local thresholding approach to flood water delineation using Sentinel-1 SAR imagery, ISPRS Journal of Photogrammetry and Remote Sensing 159: 53-62. 10.1016/j.isprsjprs.2019.10.017
7
Martinis, S., J. Kersten, and A. Twele, 2015, A fully automated TerraSAR-X based flood service, ISPRS Journal of Photogrammetry and Remote Sensing 104: 203-212. 10.1016/j.isprsjprs.2014.07.014
8
Nemni, E., J. Bullock, S. Belabbes, and L. Bromley, 2020, Fully convolutional neural network for rapid flood segmentation in synthetic aperture radar imagery, Remote Sensing 12(16): 2532. 10.3390/rs12162532
9
Sun, K., B. Xiao, D. Liu, and J. Wang, 2019, Deep high-resolution representation learning for human pose estimation, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition 5693-5703. 10.1109/CVPR.2019.00584
10
Tabari, H., 2020, Climate change impact on flood and extreme precipitation increases with water availability, Scientific Reports 10: 13768. 10.1038/s41598-020-70816-232792563PMC7426818
11
Tay, C., S. Yun, S. Chin, A. Bhardwaj, J. Jung, and E. Hill, 2020, Rapid flood and damage mapping using synthetic aperture radar in response to Typhoon Hagibis, Japan, Scientific Data 7: 100. 10.1038/s41597-020-0443-532214103PMC7096419
12
Wang, J., K. Sun, T. Cheng, B. Jiang, C. Deng, Y. Zhao, D. Liu, Y. Mu, M. Tan, X. Wang, W. Liu, and B. Xiao, 2019, Deep high-resolution representation learning for visual recognition, arXiv Reprint, arXiv:1908.07919.
13
WMO, 2022, Southeast Asia Flash Flood Guidance System Launched, https://public.wmo.int/en/media/news/southeast-asia-flash-flood-guidance-system-launched, accessed on Dec 1, 2022.
14
이도이･박소련･서동주･김진수, 2022, “UNet 기반 Sentinel-1 SAR 영상을 이용한 수체탐지: 섬진강유역 대상으로.,” 대한원격탐사학회지 38(5): 901-912.
15
이승수･이문환･강형식, 2020, “2020년 홍수 현황과 항구적 대책 방향,” KEI 포커스 8(14).
Information
  • Publisher :The Korean Association of Professional Geographers
  • Publisher(Ko) :국토지리학회
  • Journal Title :국토지리학회지
  • Journal Title(Ko) :THE GEOGRAPHICAL JOURNAL OF KOREA
  • Volume : 56
  • No :4
  • Pages :409-419
  • Received Date : 2022-12-22
  • Accepted Date : 2022-12-26
  • DOI :https://doi.org/10.22905/kaopqj.2022.56.4.8
Journal Informaiton 국토지리학회지 국토지리학회지
Online Submission
  • NRF
  • KISTI Current Status
  • KISTI Cited-by
  • KCI 문헌 유사도 검사
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close