[논문]그라운드-롤 제거를 위한 CNN과 GAN 기반 딥러닝 모델 비교 분석

조상인; 편석준

doi:10.7582/gge.2023.26.2.037

[국내논문] 그라운드-롤 제거를 위한 CNN과 GAN 기반 딥러닝 모델 비교 분석
Comparison of CNN and GAN-based Deep Learning Models for Ground Roll Suppression 원문보기

지구물리와 물리탐사 = Geophysics and geophysical exploration, v.26 no.2, 2023년, pp.37 - 51

조상인 (인하대학교 에너지자원공학과) , 편석준 (인하대학교 에너지자원공학과)

초록
AI-Helper

그라운드-롤(ground roll)은 육상 탄성파 탐사 자료에서 가장 흔하게 나타나는 일관성 잡음(coherent noise)이며 탐사를 통해 얻고자 하는 반사 이벤트 신호보다 훨씬 큰 진폭을 가지고 있다. 따라서 탄성파 자료 처리에서 그라운드-롤 제거는 매우 중요하고 필수적인 과정이다. 그라운드-롤 제거를 위해 주파수-파수 필터링, 커브릿(curvelet) 변환 등 여러 제거 기술이 개발되어 왔으나 제거 성능과 효율성을 개선하기 위한 방법에 대한 수요는 여전히 존재한다. 최근에는 영상처리 분야에서 개발된 딥러닝 기법들을 활용하여 탄성파 자료의 그라운드-롤을 제거하고자 하는 연구도 다양하게 수행되고 있다. 이 논문에서는 그라운드-롤 제거를 위해 CNN (convolutional neural network) 또는 cGAN (conditional generative adversarial network)을 기반으로 하는 세가지 모델(DnCNN (De-noiseCNN), pix2pix, CycleGAN)을 적용한 연구들을 소개하고 수치 예제를 통해 상세히 설명하였다. 알고리듬 비교를 위해 동일한 현장에서 취득한 송신원 모음을 훈련 자료와 테스트 자료로 나누어 모델을 학습하고, 모델 성능을 평가하였다. 이러한 딥러닝 모델은 현장자료를 사용하여 훈련할 때, 그라운드-롤이 제거된 자료가 필요하므로 주파수-파수 필터링으로 그라운드-롤을 제거하여 정답자료로 사용하였다. 딥러닝 모델의 성능 평가 및 훈련 결과 비교는 정답 자료와의 유사성을 기본으로 상관계수와 SSIM (structural similarity index measure)과 같은 정량적 지표를 활용하였다. 결과적으로 DnCNN 모델이 가장 좋은 성능을 보였으며, 다른 모델들도 그라운드-롤 제거에 활용될 수 있음을 확인하였다.

Abstract ▼ AI-Helper

The ground roll is the most common coherent noise in land seismic data and has an amplitude much larger than the reflection event we usually want to obtain. Therefore, ground roll suppression is a crucial step in seismic data processing. Several techniques, such as f-k filtering and curvelet transform, have been developed to suppress the ground roll. However, the existing methods still require improvements in suppression performance and efficiency. Various studies on the suppression of ground roll in seismic data have recently been conducted using deep learning methods developed for image processing. In this paper, we introduce three models (DnCNN (De-noiseCNN), pix2pix, and CycleGAN), based on convolutional neural network (CNN) or conditional generative adversarial network (cGAN), for ground roll suppression and explain them in detail through numerical examples. Common shot gathers from the same field were divided into training and test datasets to compare the algorithms. We trained the models using the training data and evaluated their performances using the test data. When training these models with field data, ground roll removed data are required; therefore, the ground roll is suppressed by f-k filtering and used as the ground-truth data. To evaluate the performance of the deep learning models and compare the training results, we utilized quantitative indicators such as the correlation coefficient and structural similarity index measure (SSIM) based on the similarity to the ground-truth data. The DnCNN model exhibited the best performance, and we confirmed that other models could also be applied to suppress the ground roll.

주제어

표/그림 (17)

그림 Fig. 1. DnCNN architecture (modified after Zhang et al., 2017).
그림 Fig. 2. Workflow of ground roll prediction using DnCNN (Li et al., 2018).
그림 Fig. 3. pix2pix architecture (modified after Isola et al., 2017).
그림 Fig. 4. Forward cycle-consistency and backward cycle-consistency (modified after Zhu et al., 2017).
그림 Fig. 5. CycleGAN architecture (modified after kaji and Kida, 2019).
표 Table 1. Hyperparameters used for training each machine learning model.
그림 Fig. 6. DnCNN model training: (a) structure of the DnCNN used, and (b) training (yellow) and validation losses (red) versus the number of epochs.
그림 Fig. 7. Structure of the GAN: (a) U-Net (Ronneberger et al., 2015) architecture used for the generator, and (b) 5 layers (3 downsampling layers plus 2 convolutional layers) for the discriminator.
그림 Fig. 8. Workflow of training the pix2pix algorithm for ground roll attenuation.
그림 Fig. 9. pix2pix model training: training (yellow) and validation losses (red) versus the number of epochs
그림 Fig. 10. Workflow of training the CycleGAN algorithm for ground roll attenuation.
그림 Fig. 11. CycleGAN model training: training (yellow) and validation losses (red) versus the number of epochs.
그림 Fig. 12. Comparison between (a) the number of trainable parameters of CNN, pix2pix and CycleGAN and, (b) elapsed time for those models to train.
그림 Fig. 13. Ground roll attenuation of the (a) 110th common shot gather. Attenuation results using (b) f-k filtering, (c) DnCNN, (d) pix2pix, and (e) CycleGAN. (f), (g), (h) and (i) are differences between original data and ground roll attenuated data: (a)-(b), (a)-(c), (a)-(d) and (a)-(e) in order.
그림 Fig. 14. F-k spectrum of (a) 110^th CSG, and ground roll attenuation results using (b) f-k filtering, (c) DnCNN, (d) pix2pix, and (e) CycleGAN.
그림 Fig. 15. Comparison of Pearson correlation coefficient between the traces in f-k filtered data and the traces in raw data (pink dotted line), DnCNN results (blue dot), pix2pix results (red dot), and CycleGAN results (green dot). The correlation coefficient of two test data, (a) and (b), are plotted.
그림 Fig. 16. Comparison of SSIM distributions using the histogram. Each SSIM is calculated between the f-k filtered data and (a) DnCNN, (b) pix2pix, and (c) CycleGAN results. The distribution of SSIM values between f -k f iltered data and raw data i s shown a s a blue histogram in all figures.

참고문헌 (35)

Benaim, S., and Wolf, L., 2017, One-sided unsupervised domain？mapping, Advances in Neural Information Processing Systems, 30. https://doi.org/10.48550/arXiv.1706.00826
Choi, W. C., Lee, G. H., Cho, S. I., Choi, B. H., and Pyun, S. J.,？2020, Fault Detection for Seismic Data Interpretation Based？on Machine Learning: Research Trends and Technological？Introduction, Geophysics and Geophysical Exploration, 23(2),？97-114 (In Korean with English abstract). https://doi.org/10.7582/GGE.2020.23.2.097

원문보기 상세보기
Deighan, A. J., and Watts, D. R., 1997, Ground-roll suppression？using the wavelet transform, Geophysics, 62(6), 1896-1903.？https://doi.org/10.1190/1.1444290

상세보기
Embree, P., Burg, J. P., and Backus, M. M., 1963, Wide-band？velocity filtering; The Pie-Slice process, Geophysics, 28(6),？948-974. https://doi.org/10.1190/1.1439310

상세보기
Fomel, S., 2002, Applications of plane-wave destruction filters,？Geophysics, 67(6), 1946-1960. https://doi.org/10.1190/1.1527095

상세보기
Fomel, S., 2009, Adaptive multiple subtraction using regularized？nonstationary regression, Geophysics, 74(1), V25-V33. https://doi.org/10.1190/1.3043447

상세보기
Gallant, E. V., Stewart, R. R., Bertram, M. B., and Lawton, D.？C., 1995, Acquisition of the Blackfoot broad-band seismic？survey, CREWES, 7(36), 1-9. https://www.crewes.org/Documents/ResearchReports/1995/1995-36.pdf
Geron, A., 2017, Hands-on machine learning with scikit-learn？and tensorflow: Concepts, Tools, and Techniques to build？intelligent systems, O'Reilly Media. https://www.oreilly.com/library/view/hands-on-machine-learning/9781491962282/
Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., and Bengio, Y., 2014,？Generative adversarial networks, arXiv preprint arXiv:1406.2661.？https://doi.org/10.48550/arXiv.1406.2661
Guo, R., Maniar, H., Di, H., Moldoveanu, N., Abubakar, A., and？Li, M., 2020, Ground roll attenuation with an unsupervised deep？learning approach, In SEG Technical Program Expanded？Abstracts 2020, Society of Exploration Geophysicists, 3164-3168. https://doi.org/10.1190/segam2020-3425792.1
He, K., Zhang, X., Ren, S., and Sun, J., 2016, Deep residual？learning for image recognition, In Proceedings of the IEEE？Conference on Computer Vision and Pattern Recognition,？770-778. https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/He_Deep_Residual_Learning_CVPR_2016_paper.pdf
Ioffe, S., and Szegedy, C., 2015, Batch normalization: Accelerating？deep network training by reducing internal covariate shift, In？International Conference on Machine Learning, 448-456.？https://doi.org/10.48550/arXiv.1502.03167
Isola, P., Zhu, J. Y., Zhou, T., and Efros, A. A., 2017, Image-to-image translation with conditional adversarial networks. In？Proceedings of the IEEE Conference on Computer Vision and？Pattern Recognition, 1125-1134. https://doi.org/10.48550/arXiv.1611.07004
Jia, Z., Lu, W., Zhang, M., and Miao, Y., 2018, Separating？ground-roll from land seismic record via convolutional neural？network, In SEG 2018 Workshop: SEG Maximizing Asset？Value Through Artificial Intelligence and Machine Learning,？60-63. DOI: 10.1190/AIML2018-16.1
Jo, Y. J., Bae, K. M., and Park, J. Y., 2020, Research Trends of？Generative Adversarial Networks and Image Generation and？Translation, Electronics and Telecommunications Trends, 35(4),？91-102 (In Korean with English abstract). https://doi.org/10.22648/ETRI.2020.J.350409

원문보기 상세보기
Kaji, S., and Kida, S., 2019, Overview of image-to-image？translation by use of deep neural networks: denoising, super-resolution, modality conversion, and reconstruction in medical？imaging, Radiological Physics and Technology, 12, 235-248.？DOI: 10.1007/s12194-019-00520-y

상세보기
Kaur, H., Fomel, S., and Pham, N., 2019, Ground roll attenuation？using generative adversarial network, 81st Annual International？Conference and Exhibition, EAGE, Extended Abstracts, 1-5.？DOI: https://doi.org/10.3997/2214-4609.201900762
Li, H., Yang, W., and Yong, X., 2018, Deep learning for ground-roll noise attenuation, In SEG Technical Program Expanded？Abstracts 2018, Society of Exploration Geophysicists, 1981-1985. https://doi.org/10.1190/segam2018-2981295.1
Liu, X., 1999, Ground roll suppression using the Karhunen-Loeve transform, Geophysics, 64(2), 564-566. https://doi.org/10.1190/1.1444562

상세보기
Liu, Y., and Fomel, S., 2013, Seismic data analysis using local？time-frequency decomposition, Geophysical Prospecting, 61(3),？516-525. DOI: 10.1111/j.1365-2478.2012.01062.x

상세보기
Mirza, M., and Osindero, S., 2014, Conditional generative？adversarial nets. arXiv preprint arXiv:1411.1784. https://doi.org/10.48550/arXiv.1411.1784
Naghizadeh, M., and Sacchi, M., 2018, Ground-roll attenuation？using curvelet downscaling, Geophysics, 83(3), V185-V195.？https://doi.org/10.1190/geo2017-0562.1

상세보기
Pham, N., and Li, W., 2022, Physics-constrained deep learning？for ground roll attenuation, Geophysics, 87(1), V15-V27.？https://doi.org/10.1190/geo2020-0691.1

상세보기
Ronneberger, O., Fischer, P., and Brox, T., 2015, U-net:？Convolutional networks for biomedical image segmentation,？In Medical Image Computing and Computer-Assisted Intervention-MICCAI 2015: 18th International Conference, Springer？International Publishing, 234-241. https://link.springer.com/chapter/10.1007/978-3-319-24574-4_28
Russell, B., Hampson, D., and Chun, J., 1990a, Noise elimination？and the Radon transform, part 1, The Leading Edge, 9(10),？18-23. https://doi.org/10.1190/1.1439677

상세보기
Russell, B., Hampson, D., and Chun, J., 1990b, Noise elimination？and the Radon transform, part 2, The Leading Edge, 9(11),？31-37. https://doi.org/10.1190/1.1439700

상세보기
Serdyukov, A. S., 2022, Ground-roll extraction using the？Karhunen-Loeve transform in the time-frequency domain,？Geophysics, 87(2), A19-A24. https://doi.org/10.1190/geo2021-0453.1

상세보기
Taigman, Y., Polyak, A., and Wolf, L., 2016, Unsupervised？cross-domain image generation, arXiv preprint arXiv:1611.02200.？https://doi.org/10.48550/arXiv.1611.02200
Treitel, S., Shanks, J. L., and Frasier, C. W., 1967, Some aspects？of fan filtering, Geophysics, 32(5), 789-800. https://doi.org/10.1190/1.1439889

상세보기
Yarham, C., Boeniger, U., and Herrmann, F., 2006, Curvelet-based ground roll removal, 76th Annual International Meeting,？SEG, Expanded Abstracts, 2777-2782. DOI: 10.1190/1.2370101
Yilmaz, O., 2001, Seismic data analysis: Processing, inversion,？and interpretation of seismic data, Society of exploration？geophysicists. https://doi.org/10.1190/1.9781560801580
Yuan, Y., Si, X., and Zheng, Y., 2020, Ground-roll attenuation？using generative adversarial networks, Geophysics, 85(4),？WA255-WA267. https://doi.org/10.1190/geo2019-0414.1

상세보기
Yuan, Y., Zhou, Z., Niu, B., Wang, H., and Xiang, A., 2005, A？method for improving the signal to noise ratio in seismic data,？Oil Geophysical Prospecting, 40, 168-171.
Zhang, K., Zuo, W., Chen, Y., Meng, D., and Zhang, L., 2017,？Beyond a gaussian denoiser: Residual learning of deep CNN？for image denoising, IEEE Transactions on Image Processing,？26(7), 3142-3155. https://doi.org/10.1109/TIP.2017.2662206

상세보기
Zhu, J. Y., Park, T., Isola, P., and Efros, A. A., 2017, Unpaired？image-to-image translation using cycle-consistent adversarial？networks, In Proceedings of the IEEE International Conference？on Computer Vision, 2223-2232. https://doi.org/10.48550/arXiv.1703.10593

저자의 다른 논문 :

표제어: PCR

동의어: Packet Collision Rate

용어 설명 출처 목록 (6)

용어 설명: PCR은 세균 특이성이 있는 primer를 이용하여 적은 수의 세균이 있을지라도 쉽게 검출할 수 있는 유용한 방법이며, 이를 이용하여 구강 내 치면세균막이나 타액에서 직접 세균을 검출할 수 있게 되었다[8].

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 논문명, 저널/프로시딩명, 저자 , 발행년, 권, 호, 시작페이지, 끝페이지, 발행기관 관리번호, 논문명, 대등논문명, 저자 , 저널/프로시딩명, 발행기관, 발행년, 발행언어, 권, 호, 시작페이지, 끝페이지, ISBN, ISSN, 주제분야, 키워드, 초록(한글), 초록(영문), 저자(소속기관)
저장형식	Text(ASCII format) Excel format RefWorks Direct Export RIS format (for Reference Manager, ProCite, EndNote), Scholar's Aids, Mendeley
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증