$\require{mediawiki-texvc}$

연합인증

연합인증 가입 기관의 연구자들은 소속기관의 인증정보(ID와 암호)를 이용해 다른 대학, 연구기관, 서비스 공급자의 다양한 온라인 자원과 연구 데이터를 이용할 수 있습니다.

이는 여행자가 자국에서 발행 받은 여권으로 세계 각국을 자유롭게 여행할 수 있는 것과 같습니다.

연합인증으로 이용이 가능한 서비스는 NTIS, DataON, Edison, Kafe, Webinar 등이 있습니다.

한번의 인증절차만으로 연합인증 가입 서비스에 추가 로그인 없이 이용이 가능합니다.

다만, 연합인증을 위해서는 최초 1회만 인증 절차가 필요합니다. (회원이 아닐 경우 회원 가입이 필요합니다.)

연합인증 절차는 다음과 같습니다.

최초이용시에는
ScienceON에 로그인 → 연합인증 서비스 접속 → 로그인 (본인 확인 또는 회원가입) → 서비스 이용

그 이후에는
ScienceON 로그인 → 연합인증 서비스 접속 → 서비스 이용

연합인증을 활용하시면 KISTI가 제공하는 다양한 서비스를 편리하게 이용하실 수 있습니다.

Sentienl-1 SAR 영상을 활용한 유류 분포특성과 CNN 구조에 따른 유류오염 탐지모델 성능 평가
Evaluation of Oil Spill Detection Models by Oil Spill Distribution Characteristics and CNN Architectures Using Sentinel-1 SAR data 원문보기

대한원격탐사학회지 = Korean journal of remote sensing, v.37 no.5 pt.3, 2021년, pp.1475 - 1490  

박소연 (서울대학교 지구환경과학전공) ,  안명환 (이화여자대학교 기후.에너지시스템공학전공) ,  이성뢰 (서울대학교 지구환경과학전공) ,  김준우 (서울대학교 지구환경과학전공) ,  전현균 (서울대학교 지구환경과학전공) ,  김덕진 (서울대학교 지구환경과학전공)

초록
AI-Helper 아이콘AI-Helper

SAR 이미지의 통계적 특징을 이용하여 유류오염영역을 특정하는 방법은 분류규칙이 복잡하고 이상값에 의한 영향을 많이 받는다는 한계가 있어, 최근 인공신경망을 기반으로 유류오염영역을 특정하는 연구가 활발히 이루어지고 있다. 하지만, 다양한 유류오염 사례에 대해 모델의 탐지 성능 및 특성을 평가한 연구는 부족하였다. 따라서, 본 연구에서는 기본적인 구조의 CNN인 Simple CNN과 픽셀 단위의 영상 분할이 가능한 U-net을 이용하여, CNN의 구조와, 유류오염의 분포특성에 따른 모델의 탐지성능차이가 존재하는지 분석하였다. 연구결과, 축소경로만 존재하는Simple CNN과 축소경로와 확장경로가 모두 존재하는U-net의 F1 score는 86.24%와 91.44%로 나타나, 두 모델 모두 비교적 높은 탐지 정확도를 보여주었지만, U-net의 탐지성능이 더 높은 것으로 나타났다. 또한 다양한 유류오염 사례에 따른 모델의 성능 비교를 위해, 유류오염의 공간적 분포특성(유류오염 주변의 육지의 분포)과 선명도(유출된 기름과 해수의 경계면이 뚜렷한 정도)를 기준으로, 유류오염 발생사례를 4가지 유형으로 구분하여 탐지 정확도를 평가하였다. Simple CNN은 각각의 유형에 대해 F1 score가 85.71%, 87.43%, 86.50%, 85.86% 로 유형별 최대 편차가 1.71%인 것으로 나타났으며, U-net은 동일한 지표에 대해 89.77%, 92.27%, 92.59%, 92.66%의 F1 score를 보여 최대 편차가 2.90% 로 두 CNN모델 모두 유류오염 분포특성에 따른 수치상 탐지성능의 차이는 크지 않은 것으로 나타났다. 하지만 모든 유류오염 유형에서 Simple CNN은 오염영역을 과대탐지 하는 경향을, U-net은 과소탐지 하는 경향을 보여, 모델의 구조와 유류오염의 유형에 따라 서로 다른 탐지 특성을 가진다는 것을 확인하였고, 이러한 특성은 유류오염과 해수의 경계면이 뚜렷하지 않은 경우 더 두드러지게 나타났다.

Abstract AI-Helper 아이콘AI-Helper

Detecting oil spill area using statistical characteristics of SAR images has limitations in that classification algorithm is complicated and is greatly affected by outliers. To overcome these limitations, studies using neural networks to classify oil spills are recently investigated. However, the st...

주제어

표/그림 (11)

AI 본문요약
AI-Helper 아이콘 AI-Helper

제안 방법

  • Groundtruth데이터는 전처리를 마친 Sentinel-1 이미지에서 유류오염 전문가의 육안판독을 통해 유류오염 영역을 1, 육지 및 선박을 포함한 비오염 영역을 0으로 라벨링하여 생성하였다. 유출영역을 정확하게 특정하는 과정에서는 전문가의 개입을 통한 digital processing 이 적용되었다.
  • 1을 따라 진행된다. Sentinel-1 영상으로부터 4단계의 전처리를 거친 후, VV 편파, VH편파, 입사각, GLCM(GreyLevelCo-occurrence Matrix) 텍스처 이미지를 이 용해학습에 이용할 입력자료를 제작하였다. 유류오염이 발생하는 다양한 환경에서 CNN의 유류오염 분석능력에 차이가 있는지를 확인하기 위해, SimpleCNN과 U-net, 두 가지 구조의 CNN을 이용해 학습하고, 각 모델의 추론 결과를 4가지 유류오염 유형으로 나누어 분석하였다.
  • 44%의 F1 score로 유류오염을 탐지하여, U- net이 더 높은 성능을 보여주었다. 또한, 유류오염 발생사례를 유류오염의 공간적 분포특성(육지의 존재 유무)과 유출된 기름과 해수경계면의 선명도에 따른 4가지 유형으로 구분하여 탐지결과를 비교하였다. 그 결과, 모든 4가지 유형에 대해 Simple CNN 모델은 오염영역을 과대 탐지하고, U-net은 과소 탐지하는 경향이 존재하였고, 이러한 경향은 유류오염의 공간적 분포 특성의 차이보다는 유류오염과 해수의 경계가 분명하지 않는 유형에서 더 뚜렷하게 나타났다.
  • 이러한 관점에서 본 연구는 발생가능한 유류오염 사례와 CNN의 구조에 따른 유류오염 탐지 성능 및 특성을 비교하는 의의를 가진다. 본 연구에서는 2가지 구조의 CNN을 이용하고, 유류오염을 4가지 유형으로 구분하여 모델의 탐지 성능을 정량적(Table2), 정성적(Fig. 8) 평가를 통해 검증하였다.
  • 본 연구에서는 이러한 통계적 방법의 한계를 극복하고, CNN의 유류오염 탐지성능이 다양한 오염유형에 대하여 어떠한 차이를 보이는지 분석하기 위해, 1) 두 가지 구조의 CNN을 이용하여 유류오염 영역과 비오염영역을 구분하고, 2) CNN 의 구조에 따른 성능의 차이가 있는지, 3) CNN이 유류오염이 발생하는 다양한 경우에 있어서 탐지성능의 차이를 보이는지, 유류오염을 4가지 유형으로 나누어 분석하였다
  • 연구 방법은 총 4개의 절로 나누어 2장에 제시되었으며, 각각의 절에서는 Sentine-1 자료취득 방법 및 전처리기법, groundtruth데이터 구축방법, 연구에 이용된 CNN구조와 hyperparameter설정, 마지막으로 모델의 정확도 검증 방법에 대하여 설명하였다. 3장에서는 테스트 이미지에 대해, 연구에서 이용된 2개의 CNN모델이 추론한 결과를 앞서 구분한 4가지 유류오염 유형에 따라 해석하였다.
  • Sentinel-1 영상으로부터 4단계의 전처리를 거친 후, VV 편파, VH편파, 입사각, GLCM(GreyLevelCo-occurrence Matrix) 텍스처 이미지를 이 용해학습에 이용할 입력자료를 제작하였다. 유류오염이 발생하는 다양한 환경에서 CNN의 유류오염 분석능력에 차이가 있는지를 확인하기 위해, SimpleCNN과 U-net, 두 가지 구조의 CNN을 이용해 학습하고, 각 모델의 추론 결과를 4가지 유류오염 유형으로 나누어 분석하였다.

대상 데이터

  • Simple CNN과 U-net모델의 훈련과 성능평가에는 동일한 훈련 데이터셋과 테스트 데이터셋을 사용하였다. 총 122개의 Sentinel-1 SAR 유류오염이미지 중 약 80%인 100개의 이미지는 다시 80%와 20%로 나뉘어 각각 모델 학습과 검증을 위한 자료로 이용하였고, 전체 자료에서 무작위로 선택된 나머지 20%인 22개 이미지는 학습된 모델을 이용하여 추론하는데 사용하였다.

데이터처리

  • 또한, 해상 유류오염 영상 내 육지는, 해수면과 비교하여 산란 특성이 매우 다르고, 비연속적으로 존재하기 때문에 영상 내 육지의 존재 유무 역시 영상 분류 정확도에 영향을 미칠 수 있다고 가정하였다. 각각의 유형에 따른 모델의 추론 결과는 정량적으로 평가되었고, 정량적 평가에서 나타나지 않는 유류오염 특성별 분류결과를 육안 판독을 통해 비교하였다. 유류 오염 분포 특성에 따른 4가지 유형은 다음과 같다.

이론/모형

  • 두 분류모델의 성능을 정량적으로 평가하기 위해, 픽셀 단위의 분류모델 성능평가지표(confusionmatrix)를 이용하였다(Table1).
  • 마지막으로, 정밀도와 재현율은 서로 상충하는 개념의 지표이기 때문에 두 지표의 조화 평균을 나타내는 F1 score를 이용하여 모델의 전반적인 성능을 하나의 수로 표현할 수 있다. 정확도 검증을 위한 4가지 지표는 모델이 추론한 이미지와 groundtruth 이미지를 픽셀단위로 비교하여 confusion matrix를 생성하고, 이를 수식에 대입하여 계산되었다.
본문요약 정보가 도움이 되었나요?

참고문헌 (33)

  1. Alom, M.Z., T.M. Taha, C. Yakopcic, S. Westberg, P. Sidike, M.S. Nasrin, B.C. Van Esesn, A.A.S. Awwal, and V.K. Asari, 2018. The history began from alexnet: A comprehensive survey on deep learning approaches, arXiv preprint, arXiv (1803.01164): 1-39. 

  2. Alpers, W. and H. Huhnerfuss, 1988. Radar signatures of oil films floating on the sea surface and the Marangoni effect, Journal of Geophysical Research: Oceans, 93(C4): 3642-3648. 

  3. Angelliaume, S., P.C. Dubois-Fernandez, C.E. Jones, B. Holt, B. Minchew, E. Amri, and V. Miegebielle, 2018. SAR imagery for detecting sea surface slicks: Performance assessment of polarization-dependent parameters, IEEE Transactions on Geoscience and Remote Sensing, 56(8): 4237-4257. 

  4. Bjerde, K.W., A.H. Solberg, and R. Solberg, 1993. Oil spill detection in SAR imagery, Proc. of 1993 International Geoscience and Remote Sensing Symposium, Tokyo, JP, Aug. 18-21, vol. 3, pp. 943-945. 

  5. Brekke, C. and A.H. Solberg, 2005. Oil spill detection by satellite remote sensing, Remote Sensing of Environment, 95(1): 1-13. 

  6. De Boer, P.T., D.P. Kroese, S. Mannor, and R.Y. Rubinstein, 2005. A tutorial on the cross-entropy method, Annals of Operations Research, 134(1): 19-67. 

  7. De Souza, D.L., A.D. Neto, and da W. Mata, 2006. Intelligent system for feature extraction of oil slick in sar images: Speckle filter analysis, Proc. of the 13th international conference on Neural Information Processing, Hong Kong, CN, Oct. 3-6, vol. 2, pp. 729-736. 

  8. Del Frate, F., A. Petrocchi, J. Lichtenegger, and G. Calabresi, 2000. Neural networks for oil spill detection using ERS-SAR data, IEEE Transactions on Geoscience and Remote Sensing, 38(5): 2282-2287. 

  9. Fan, J., F. Zhang, D. Zhao, and J. Wang, 2015. Oil Spill Monitoring Based on SAR Remote Sensing Imagery, Aquatic Procedia, 3: 112-118. 

  10. Filipponi, F., 2019. Sentinel-1 GRD preprocessing workflow, Proc. of 2019 3rd International Electronic Conference on Remote Sensing, Roma, IT, May. 22-June. 5, vol. 18, p. 11. 

  11. Fingas, M. and C. Brown, 2014. Review of oil spill remote sensing, Marine Pollution Bulletin, 83(1): 9-23. 

  12. Fiscella, B., A. Giancaspro, F. Nirchio, P. Pavese, and P. Trivero, 2000. Oil spill detection using marine SAR images, International Journal of Remote Sensing, 21(18): 3561-3566. 

  13. Glorot, X., A. Bordes, and Y. Bengio, 2011. Deep sparse rectifier neural networks, Proc. of the fourteenth international conference on artificial intelligence and statistics, Fort Lauderdale, FI, Apr. 11-13, vol. 15, pp. 315-323. 

  14. Hammoud, B., F. Ndagijimana, G. Faour, H. Ayad, and J. Jomaah, 2019. Bayesian statistics of wideband radar reflections for oil spill detection on rough ocean surface, Journal of Marine Science and Engineering, 7(1): 12. 

  15. Haralick, R.M., K. Shanmugam, and I.H. Dinstein, 1973. Textural Features for Image Classification, IEEE Transactions on Systems, Man, and Cybernetics, SMC-3(6): 610-621. 

  16. Hui, K.P., N. Bean, M. Kraetzl and D.P. Kroese, 2005. The cross-entropy method for network reliability estimation, Annals of Operations Research, 134(1): 101-118. 

  17. Kingma, D.P. and J. Ba, 2014. Adam: A method for stochastic optimization, arXiv preprint, arXiv (1412.6980): 1-15. 

  18. Krestenitis, M., G. Orfanidis, K. Ioannidis, K. Avgerinakis, S. Vrochidis, and I. Kompatsiaris, 2019. Oil spill identification from satellite images using deep neural networks, Remote Sensing, 11(15): 1762. 

  19. Lee, J.S., J.H. Wen, T.L. Ainsworth, K.S. Chen, and A.J. Chen, 2008. Improved sigma filter for speckle filtering of SAR imagery, IEEE Transactions on Geoscience and Remote Sensing, 47(1): 202-213. 

  20. Lu, J., 2003. Marine oil spill detection, statistics and mapping with ERS SAR imagery in south-east Asia, International Journal of Remote Sensing, 24(15): 3013-3032. 

  21. Marghany, M., 2001. RADARSAT automatic algorithms for detecting coastal oil spill pollution, International Journal of Applied Earth Observation and Geoinformation, 3(2): 191-196. 

  22. Miranda, N., P.J. Meadows, D. Type, and T. Note, 2015. Radiometric calibration of S-1 Level-1 products generated by the S-1 IPF, https://sentinel.esa.int/documents/247904/685163/S1-RadiometricCalibration-V1.0.Pdf, Accessed on Sep. 7, 2021. 

  23. Nirchio, F., M. Sorgente, A. Giancaspro, W. Biamino, E. Parisato, R. Ravera, and P. Trivero, 2005. Automatic detection of oil spills from SAR images, International Journal of Remote Sensing, 26(6): 1157-1174. 

  24. Ronneberger, O., P. Fischer, and T. Brox, 2015. U-net: Convolutional networks for biomedical image segmentation, Proc. of the International Conference on Medical image computing and computer-assisted intervention, Munich, DE, Oct. 5-9, pp. 234-241. 

  25. Sheta, A., M. Alkasassbeh, M. Braik, and H.A. Ayyash, 2012. Detection of oil spills in SAR images using threshold segmentation algorithms, International Journal of Computer Applications, 57(7): 10-15. 

  26. Singha, S., T.J. Bellerby, and O. Trieschmann, 2013. Satellite oil spill detection using artificial neural networks, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 6(6): 2355-2363. 

  27. Solberg, A.S., G. Storvik, R. Solberg, and E. Volden, 1999. Automatic detection of oil spills in ERS SAR images, IEEE Transactions on Geoscience and Remote Sensing, 37(4): 1916-1924. 

  28. Stathakis, D., K. Topouzelis, and V. Karathanassi, 2006. Large-scale feature selection using evolved neural networks, Proc. of Image and Signal Processing for Remote Sensing XII, Stockholm, SE, Sep. 11-14, vol. 6365, pp. 636513.1-636513.9. 

  29. Tharwat, A., 2020. Classification assessment methods, Applied Computing and Informatics, 17(1): 168-192. 

  30. Topouzelis, K., V. Karathanassi, P. Pavlakis, and D. Rokos, 2007. Detection and discrimination between oil spills and look-alike phenomena through neural networks, ISPRS Journal of Photogrammetry and Remote Sensing, 62(4): 264-270. 

  31. Trivero, P., B. Fiscella, F. Gomez, and P. Pavese, 1998. SAR detection and characterization of sea surface slicks, International Journal of Remote Sensing, 19(3): 543-548. 

  32. Yekeen, S.T., A.L. Balogun, and K.B.W. Yusof, 2020. A novel deep learning instance segmentation model for automated marine oil spill detection, ISPRS Journal of Photogrammetry and Remote Sensing, 167: 190-200. 

  33. Zeng, L., M. Schmitt, L. Li, and X.X. Zhu, 2017. Analysing changes of the Poyang Lake water area using Sentinel-1 synthetic aperture radar imagery, International Journal of Remote Sensing, 38(23): 7041-7069. 

저자의 다른 논문 :

관련 콘텐츠

오픈액세스(OA) 유형

GOLD

오픈액세스 학술지에 출판된 논문

저작권 관리 안내
섹션별 컨텐츠 바로가기

AI-Helper ※ AI-Helper는 오픈소스 모델을 사용합니다.

AI-Helper 아이콘
AI-Helper
안녕하세요, AI-Helper입니다. 좌측 "선택된 텍스트"에서 텍스트를 선택하여 요약, 번역, 용어설명을 실행하세요.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.

선택된 텍스트

맨위로