[논문]U-Net과 cWGAN을 이용한 탄성파 탐사 자료 보간 성능 평가

유지윤; 윤대웅

doi:10.7582/gge.2022.25.3.140

U-Net과 cWGAN을 이용한 탄성파 탐사 자료 보간 성능 평가
Comparison of Seismic Data Interpolation Performance using U-Net and cWGAN 원문보기

지구물리와 물리탐사 = Geophysics and geophysical exploration, v.25 no.3, 2022년, pp.140 - 161

유지윤 (전남대학교 에너지자원공학과) , 윤대웅 (전남대학교 에너지자원공학과)

초록
AI-Helper

탄성파 탐사 자료 획득 시 자료의 일부가 손실되는 문제가 발생할 수 있으며 이를 위해 자료 보간이 필수적으로 수행된다. 최근 기계학습 기반 탄성파 자료 보간법 연구가 활발히 진행되고 있으며, 특히 영상처리 분야에서 이미지 초해상화에 활용되고 있는 CNN (Convolutional Neural Network) 기반 알고리즘과 GAN (Generative Adversarial Network) 기반 알고리즘이 탄성파 탐사 자료 보간법으로도 활용되고 있다. 본 연구에서는 손실된 탄성파 탐사 자료를 높은 정확도로 복구하는 보간법을 찾기 위해 CNN 기반 알고리즘인 U-Net과 GAN 기반 알고리즘인 cWGAN (conditional Wasserstein Generative Adversarial Network)을 탄성파 탐사 자료 보간 모델로 사용하여 성능 평가 및 결과 비교를 진행하였다. 이때 예측 과정을 Case I과 Case II로 나누어 모델 학습 및 성능 평가를 진행하였다. Case I에서는 규칙적으로 50% 트레이스가 손실된 자료만을 사용하여 모델을 학습하였고, 생성된 모델을 규칙/불규칙 및 샘플링 비율의 조합으로 구성된 총 6가지 테스트 자료 세트에 적용하여 모델 성능을 평가하였다. Case II에서는 6가지 테스트 자료와 동일한 형식으로 샘플링된 자료를 이용하여 해당 자료별 모델을 생성하였고, 이를 Case I과 동일한 테스트 자료 세트에 적용하여 결과를 비교하였다. 결과적으로 cWGAN이 U-Net에 비해 높은 정확도의 예측 성능을 보였으며, 정량적 평가지수인 PSNR과 SSIM에서도 cWGAN이 높은 값이 나타나는 것을 확인하였다. 하지만 cWGAN의 경우 예측 결과에서 추가적인 잡음이 생성되었으며, 잡음을 제거하고 정확도를 개선하기 위해 앙상블 작업을 수행하였다. Case II에서 생성된 cWGAN 모델들을 이용하여 앙상블을 수행한 결과, 성공적으로 잡음이 제거되었으며 PSNR과 SSIM 또한 기존의 개별 모델 보다 향상된 결과를 나타내었다.

Abstract ▼ AI-Helper

Seismic data with missing traces are often obtained regularly or irregularly due to environmental and economic constraints in their acquisition. Accordingly, seismic data interpolation is an essential step in seismic data processing. Recently, research activity on machine learning-based seismic data interpolation has been flourishing. In particular, convolutional neural network (CNN) and generative adversarial network (GAN), which are widely used algorithms for super-resolution problem solving in the image processing field, are also used for seismic data interpolation. In this study, CNN-based algorithm, U-Net and GAN-based algorithm, and conditional Wasserstein GAN (cWGAN) were used as seismic data interpolation methods. The results and performances of the methods were evaluated thoroughly to find an optimal interpolation method, which reconstructs with high accuracy missing seismic data. The work process for model training and performance evaluation was divided into two cases (i.e., Cases I and II). In Case I, we trained the model using only the regularly sampled data with 50% missing traces. We evaluated the model performance by applying the trained model to a total of six different test datasets, which consisted of a combination of regular, irregular, and sampling ratios. In Case II, six different models were generated using the training datasets sampled in the same way as the six test datasets. The models were applied to the same test datasets used in Case I to compare the results. We found that cWGAN showed better prediction performance than U-Net with higher PSNR and SSIM. However, cWGAN generated additional noise to the prediction results; thus, an ensemble technique was performed to remove the noise and improve the accuracy. The cWGAN ensemble model removed successfully the noise and showed improved PSNR and SSIM compared with existing individual models.

주제어

표/그림 (26)

그림 Fig. 1. U-Net architecture (modified after Ronneberger et al., 2015).
그림 Fig. 2. cWGAN architecture.
그림 Fig. 3. Differences between MSE and perceptual loss on manifold (modified after Ledig et al., 2017).
그림 Fig. 4. Bagging predictors.
그림 Fig. 6. Input data for the six cases. (a) is the label data. (b), (c), and (d) are the input data with regular missing traces corresponding to 50%, 67%, and 75%, respectively, (e), (f), and (g) are input data with irregular missing traces corresponding to 50%, 70%, and 80%, respectively.
그림 Fig. 5. Acquisition geometry in the SEG/EAGE salt model. Sourcereceiver configuration of a shot. The thick line is the streamer line that is chosen as a near offset data (modified after Ji and Choi, 2010).
표 Table 1. SEG/EAGE salt model geometry.
그림 Fig. 7. Two cases used for data prediction. Diagrams of (a) Case I and (b) Case II.
그림 Fig. 8. Architecture of the U-Net used.
그림 Fig. 9. Structure of the cWGAN used. The label data and input data are y and x, respectively. The generator and discriminator are G and D, respectively. D is able to predict the output generated by G from the target (modified after Wei et al., 2021).
그림 Fig. 10. Test results for the six cases input data in Case I. Interpolation results using (a) U-Net and (b) cWGAN. (c) and (d) are the residuals for (a) and (b), respectively.
그림 Fig. 10. Continued
그림 Fig. 11. Test results for the six cases input data in Case II. Interpolation results using (a) U-Net and (b) cWGAN. (c) and (d) are the residuals for (a) and (b), respectively.
그림 Fig. 11. Continued
그림 Fig. 12. Results of f-k spectrum from the interpolation results of data with irregularly 80% missing traces in Fig. 10. (a) Input data, (b) label data, and (c) interpolation result using U-Net. (d) Residual between (b) and (c), (e) interpolation result using cWGAN, and (f) residual between (b) and (e).
그림 Fig. 13. A selected trace for comparison. Traces for six input data in (a) Case I and (b) Case II.
그림 Fig. 13. Continued
그림 Fig. 14. Comparison of PSNR and SSIM by the algorithm used in Case I and Case II. Input data with (a) regularly and (b) irregularly missing traces in Case I. Input data with (c) regularly and (d) irregularly missing traces in Case II.
그림 Fig. 15. Test results for the six cases input data in the ensemble. Interpolation results using (a) U-Net and (b) cWGAN. (c) and (d) are the residuals for (a) and (b), respectively.
그림 Fig. 15. Continued
그림 Fig. 16. Comparison of interpolation results using cWGAN for data of irregularly 80% missing traces. (a) Label data and interpolation results in (b) Case I, (c) Case II, and (d) the ensemble.
그림 Fig. 17. Results of f-k spectrum from the interpolation results of data with irregularly 80% missing traces in Fig. 15. (a) Input data, (b) label data, and (c) interpolation result using U-Net. (d) Residual between (b) and (c), (e) interpolation result using cWGAN, and (f) residual between (b) and (e).
그림 Fig. 18. Selected trace for comparison of test results using the ensemble model.
그림 Fig. 19. Comparison of PSNR and SSIM from the test results using the ensemble model. Input data with (a) regularly and (b) irregularly missing traces.
그림 Fig. 20. Comparisons between PSNR of U-Net and cWGANs for Case I, Case II, and the ensemble. Data with (a) regularly and (b) irregularly missing traces using the U-Net. Data with (c) regularly and (d) irregularly missing traces using the cWGAN.
그림 Fig. 21. Comparison between SSIM of U-Net and cWGANs for Case I, Case II, and the ensemble. Data with (a) regularly and (b) irregularly missing traces using the U-Net. Data with (c) regularly and (d) irregularly missing traces using the cWGAN.

참고문헌 (29)

Arjovsky, M., Chintala, S., and Bottou, L., 2017, Wasserstein Generative Adversarial Networks, Proceedings of the 34th International Conference on Machine Learning, PMLR 70, 214-223. https://proceedings.mlr.press/v70/arjovsky17a.html
Bae, W., Kwon, Y., and Ha, W., 2020, Research Trend analysis for Seismic Data Interpolation Methods using Machine Learning, Geophysics and Geophysical Exploration, 23(3), 192-207. https://doi.org/10.7582/GGE.2020.23.3.00192

원문보기 상세보기
Bauer, E., and Kohavi, R., 1999, An empirical comparison of voting classification algorithm: bagging, boosting, and variants, Machine Learning, 36(1-2), 105-142. https://link.springer.com/article/10.1023/A:1007515423169
Breiman, L., 1996, Bagging predictors, Machine Learning, 24, 123-140. https://link.springer.com/article/10.1007/BF00058655

상세보기
Choi, J., Byun, J., and Seol, S. J., 2014, Wavelet Based Matching Pursuit Method for Interpolation of Seismic Trace with Spatial Aliasing, Geophysics and Geophysical Exploration, 17(2), 88-94. https://doi.org/10.7582/GGE.2014.17.2.088

원문보기 상세보기
Dietterich, T. G., 2000a, An experimental comparison of decision trees: bagging, boosting, and randomization, Machine Learning, 40(2), 139-157. https://link.springer.com/article/10.1023/A:1007607513941
Dietterich, T. G., 2000b, Ensemble method in machine learning, LNCS, 1857, 1-15. https://link.springer.com/chapter/10.1007/3-540-45014-9_1

상세보기
Isola, P., Zhu, J. Y., Zhou, T., and Efros, A. A., 2017, Image-to-Image Translation with Conditional Adversarial Networks, IEEE conference on computer vision and pattern recognition, 1125-1134. https://openaccess.thecvf.com/content_cvpr_2017/html/Isola_Image-To-Image_Translation_With_CVPR_2017_paper.html
Ji, J. and Choi, Y. G., 2010, 3D Seismic Data Processing Methodology using Public Domain Software System, Geophysics and Geophysical Exploration, 13(2), 159-168. https://koreascience.kr/article/JAKO201026359284114.page
Kim, B., Jeong, S., and Byun, J., 2012, Curvelet transform-based POCS in f-k domain, SEG Expanded Abstracts of the 82nd Annual International Meeting, 1-5. https://doi.org/10.1190/segam2012-1063.1
Kwak, S., and Kim, H., 2014, Comparison of ensemble pruning methods using Lasso-bagging and WAVE-bagging, Journal of the Korean Data & Information Science Society, 25(6), 1371-1383. https://doi.org/10.7465/jkdi.2014.25.6.1371

원문보기 상세보기
Ledig, C., Theis, L., Huszar, F., Caballero, J., Cunningham, A., Acosta, A., Aitken, A., Tejani, A., Totz, J., Wang, Z., and Shi, W., 2017, Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network, Computer Vision and Pattern Recognition, 5, 4681-4690. https://openaccess.thecvf.com/content_cvpr_2017/html/Ledig_Photo-Realistic_Single_Image_CVPR_2017_paper.html
Lee, B., Yang, J., and Kim, S., 2009, Ensemble Learning of Regional Experts, Journal of KIISE, 15(2), 135-139. https://koreascience.kr/article/JAKO200907841292048.page
Liu, B., and Sacchi, M. D., 2004, Minimum weighted norm interpolation of seismic records, Geophysics, 69(6), 1560-1568. https://doi.org/10.1190/1.1836829

상세보기
Naghizadeh, M., and Sacchi, M. D., 2010, On sampling functions and Fourier reconstruction methods, Geophysics, 75(6), WB137-WB151. https://doi.org/10.1190/1.3503577

상세보기
Mirza, M., and Osindero, S., 2014, Conditional Generative Adversarial Nets, arXiv preprint arXiv:1411.1784. https://doi.org/10.48550/arXiv.1411.1784
Optiz, D., and Maclin, R., 1999, Popular ensemble methods: an empirical study, Journal of AIR, 11, 169-198. https://doi.org/10.1613/jair.614

상세보기
Porsani, M. J., 1999, Seismic trace interpolation using half-step prediction filters, Geophysics, 64(5), 1461-1467. https://doi.org/10.1190/1.1444650

상세보기
Ronneberger, O., Fischer, P., and Brox, T., 2015, U-Net: Convolutional Networks for Biomedical Image Segmentation, Medical Image Computing and Computer-Assisted Intervention, 9351, 1-8. https://link.springer.com/chapter/10.1007/978-3-319-24574-4_28
Schonewille, M., Klaedtke, A., and Vigner, A., 2009, Anti-alias anti-leakage Fourier transform, SEG Expanded Abstracts of the 79th Annual International Meeting, 3249-3253. https://onepetro.org/SEGAM/proceedings-abstract/SEG09/All-SEG09/SEG-2009-3249/97241
Siahkoohi, A., Kumar, R., and Herrmann, F., 2018, Seismic Data Reconstruction with Genenrative Adversarial Networks, 80th EAGE Conference and Exhibition, 2018, 1-5. https://doi.org/10.3997/2214-4609.201801393
Spitz, S., 1991, Seismic trace interpolation in the F-X domain, Geophysics, 56(6), 785-794. https://doi.org/10.1190/1.1443096

상세보기
Wang, B., Zhang, N., Lu, W., and Wang, J., 2019, Deep-learning-based seismic data interpolation: A preliminary result, Geophysics, 84(1), V11-V20. https://doi.org/10.1190/geo2017-0495.1

상세보기
Wang, Y., Wang, B., Tu, N., and Geng, J., 2020, Seismic trace interpolation for irregularly spatial sampled data using convolutional autoencoder, Geophysics, 85(2), V119-V130. https://doi.org/10.1190/geo2018-0699.1

상세보기
Wang, Z., Bovik, A. C., Sheikh, H. R., and Simoncelli, E. P., 2004, Image Quality Assessment: From Error Visibility to Structural Similarity, IEEE Transaction on Image Processing, 13(4), 600-612. https://ieeexplore.ieee.org/abstract/document/1284395
Wei, Q., Li, X. Y., and Song, M. P., 2021, De-aliased seismic data interpolation using conditional Wasserstein generative adversarial networks, Computer and Geosciences, 154, 1-13. https://doi.org/10.1016/j.cageo.2021.104801

상세보기
Xu, S., Zhang, Y., and Lambare, G., 2010, Antileakage Fourier transform for seismic data regularization in higher dimentions, Geophysics, 75, WB113-WB120. https:doi.org/10.1190/1.1993713

상세보기
Yang, A. Y., and Suh, J. H., 2003, Applying Spitz Trace Interpolation Algorithm for Seismic Data, Geophysics, 6(4), 171-179. https://koreascience.kr/article/JAKO200307921810193.page
Yang, P., Gao, J., and Chen, W., 2012, Curvelet-based POCS interpolation of nonuniformly sampled seismic records, Journal of Applied Geophysics, 79, 90-99. https://doi.org/10.1016/j.jappgeo.2011.12.004

상세보기

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