Deep Learning-based Diffraction Separation Considering Time-Variant Characteristics

Junyon Park; Evan Um; Sooyoon Kim; SoonJee Seol; Joongmoo Byun

doi:10.7582/GGE.2026.29.2.100

All Issue

2026 Vol.29, Issue 2 Preview Page Next Page

Research Article

Deep Learning-based Diffraction Separation Considering Time-Variant Characteristics 시변 특성을 고려한 딥러닝 기반 회절 분리

31 May 2026. pp. 100-111

PDF XML

Abstract

Diffractions in seismic surveys provide high-resolution information on subsurface heterogeneities and structures within reflection shadow zones. However, their small amplitudes and overlap with reflections make extraction difficult. Even recent deep learning (DL)-based methods suffer from mis-separation if they fail to account for frequency attenuation and eccentricity variations over travel time. In this study, we improved the DL algorithm to reflect time-dependent signal characteristics. By adding a time-indicator channel, the model learns varying dominant frequencies and diffraction eccentricities across different time regions. For efficient data generation, the Kirchhoff diffraction formula was used instead of the finite-difference method. Application to the Mobil AVO Viking Graben Line 12 field data showed that misidentification—specifically late-time high-frequency noise and early-time low-frequency reflections—was significantly mitigated. Furthermore, the phase and amplitude of diffractions were more accurately preserved. This method is expected to be effective for seismic data with rapidly changing frequency characteristics or deep subsurface targets.

Keywords

Diffraction Extraction

Diffraction Imaging

Time-Indicator

Kirchhoff Diffraction Formula

탄성파 탐사에서 회절파(diffraction)는 지하 불균질성 및 미세 구조에서 발생하여 고해상도로 정보를 제공하며, 반사파 음영대 내 구조 탐지도 가능하게 하는 중요한 신호이다. 그러나 진폭이 작고 반사파와 중첩되어 전통적 기법으로는 분리에 어려움이 있다. 최근의 딥러닝(deep learning) 기반 회절파 추출 방법도 취득 시간에 따른 주파수 감쇠 및 이심률(eccentricity) 변화를 제대로 반영하지 못하면 오분리 문제가 발생할 수 있다. 본 연구에서는 기존의 딥러닝 알고리즘을 시간대별 신호 특성을 반영할 수 있도록 개선하였다. 입력 자료에 시간 지시자(time-indicator) 채널을 추가해 모델이 시간 대역별로 다른 주파수 및 이심률을 학습하도록 하였으며, 효율적인 대량 인공 회절파 생성을 위해 유한차분법 대신 Kirchhoff 회절파 공식을 도입하였다. 제안 방법을 Mobil AVO Viking Graben Line 12 현장 자료에 적용한 결과, 기존 방법의 후기 시간대 고주파수 잡음 및 초기 시간대 저주파수 반사파 오인 추출 문제가 크게 완화되었으며, 회절파의 위상과 진폭이 더 정확히 보존되었다.

키워드

회절파 분리

회절파 영상화

시간 지시자

Kirchhoff 회절파 공식

References

Dafni, R. and Symes, W. W., 2017, Diffraction imaging by prestack reverse-time migration in the dip-angle domain, Geophys. Prospect., 65(S1), 295-316.

10.1111/1365-2478.12580

Dell, S. and Gajewski, D., 2011, Common-reflection-surface-based workflow for diffraction imaging, Geophysics, 76(5), S187-S195.

10.1190/geo2010-0229.1

Hilterman, F. J., 1970, Three-dimensional seismic modeling, Geophysics, 35(6), 1020-1037.

10.1190/1.1440140

Khaidukov, V., Landa, E., and Moser, T. J., 2004, Diffraction imaging by focusing-defocusing: An outlook on seismic superresolution, Geophysics, 69(6), 1478-1490.

10.1190/1.1836821

Kim, S., Seol, S. J., Byun, J., and Oh, S., 2022, Extraction of diffractions from seismic data using convolutional U-net and transfer learning, Geophysics, 87(2), V117-V129.

10.1190/geo2020-0847.1

Klokov, A. and Fomel, S., 2012, Separation and imaging of seismic diffractions using migrated dip-angle gathers, Geophysics, 77(6), S131-S143.

10.1190/geo2012-0017.1

Kozlov, E., Barasky, N., Korolev, E., Antonenko, A., and Koshchuk, E., 2004, Imaging scattering objects masked by specular reflections, 74th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1131-1134.

10.1190/1.1851082

Lowney, B., Lokmer, I., and O’Brien, G. S., 2021, Multi-domain diffraction identification: A supervised deep learning technique for seismic diffraction classification, Comput. Geosci., 155, 104845.

10.1016/j.cageo.2021.104845

Lowney, B., Lokmer, I., Bean, C. J., O’Brien, G. S., and Igoe, M., 2018, An outlook on seismic diffraction imaging using pattern recognition, 80th Annual International Conference and Exhibition, EAGE, Extended Abstracts.

10.3997/2214-4609.201800634

Moser, T. J. and Howard, C. B., 2008, Diffraction imaging in depth, Geophys. Prospect., 56(5), 627-641.

10.1111/j.1365-2478.2007.00718.x

Schneider, W. A., 1978, Integral formulation for migration in two and three dimensions, Geophysics, 43(1), 49-76.

10.1190/1.1440828

Schwarz, B. and Gajewski, D., 2017, Accessing the diffracted wavefield by coherent subtraction, Geophys. J. Int., 211(1), 45-49.

10.1093/gji/ggx291

Shin, J., Ha, J., and Lim, K., 2025, 3D diffraction imaging for ultra-high-resolution seismic data in nearshore area, Geophys. and Geophys. Explor., 28(1), 17-26 (in Korean with English abstract).

Shustak, M., and Landa, E., 2018, Time reversal for wave refocusing and scatterer detection using machine learning, Geophysics, 83(5), T257-T263.

10.1190/geo2017-0679.1

Trorey, A. W., 1970, A simple theory for seismic diffractions, Geophysics, 35(5), 762-784.

10.1190/1.1440129

Waheed, U. B., Alkhalifah, T., and Stovas, A., 2013, Diffraction traveltime approximation for TI media with an inhomogeneous background, Geophysics, 78(5), WC103-WC111.

10.1190/geo2012-0413.1

Wang, H., Liu, X., and Chen, Y., 2020, Separation and imaging of seismic diffractions using a localized rank-reduction method with adaptively selected ranks, Geophysics, 85(6), V497-V506.

10.1190/geo2020-0215.1

Information

Publisher :Korean Society of Earth and Exploration Geophysicists
Publisher(Ko) :한국지구물리물리탐사학회
Journal Title :Geophysics and Geophysical Exploration
Journal Title(Ko) :지구물리와 물리탐사
Volume : 29
No :2
Pages :100-111
Received Date : 2026-02-24
Revised Date : 2026-04-17
Accepted Date : 2026-05-17
DOI :https://doi.org/10.7582/GGE.2026.29.2.100

Geophysics and Geophysical Exploration ISSN:1229-1064(Print) 2384-051X(Online) 지구물리와 물리탐사

All Issue