Abstract:PreviousNext No AccessSEG Technical Program Expanded Abstracts 2020Ground roll attenuation with an unsupervised deep learning approachAuthors: Rui GuoHiren ManiarHaibin DiNick MoldoveanuAria AbubakarMaokun LiRui GuoSchlumbergerSearch for more papers by this author, Hiren ManiarSchlumbergerSearch for more papers by this author, Haibin DiSchlumbergerSearch for more papers by this author, Nick MoldoveanuSchlumbergerSearch for more papers by this author, Aria AbubakarSchlumbergerSearch for more papers by this author, and Maokun LiTsinghua UniversitySearch for more papers by this authorhttps://doi.org/10.1190/segam2020-3425792.1 SectionsSupplemental MaterialAboutPDF/ePub ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InRedditEmail AbstractWe propose an unsupervised deep learning approach for seismic ground roll attenuation. This method consists of two stages. While the first stage aims at detecting the first arrivals and masking the non-signal and signal portions (before and after first arrival) by a convolutional encoder-decoder network, the second stage is to decompose a seismic record into ground rolls and reflection signals by a 2D convolutional neural network (CNN) consisting of two sub-networks, which outputs the ground roll and reflection components, respectively. The criterion of such unsupervised separation is formulated in both the frequency and time domains, with the former to maximize the difference between the spectrum distribution of ground rolls and that of reflections and the latter to minimize the residual between the summation of the two sub-network outputs and the input seismic. Test results on real seismic records demonstrate that not only the proposed method is computational efficient, but more importantly the power spectra of the ground rolls are largely suppressed while preserving the reflections’ low frequency components.Presentation Date: Wednesday, October 14, 2020Session Start Time: 8:30 AMPresentation Time: 11:00 AMLocation: 360DPresentation Type: OralKeywords: machine learning, noise, rayleigh wave, processingPermalink: https://doi.org/10.1190/segam2020-3425792.1FiguresReferencesRelatedDetailsCited byPhysics-constrained deep learning for ground roll attenuationNam Pham and Weichang Li18 November 2021 | GEOPHYSICS, Vol. 87, No. 1Ground-Roll Attenuation Using a Dual-Filter-Bank Convolutional Neural NetworkIEEE Transactions on Geoscience and Remote Sensing, Vol. 60Physics-guided self-supervised learning for monochromatic noise removalYuan Zi, Wenyi Hu, Jiefu Chen, Xuqing Wu, and Zhu Han1 September 2021Physics-constrained deep learning for ground-roll attenuationNam Pham and Weichang Li1 September 2021 SEG Technical Program Expanded Abstracts 2020ISSN (print):1052-3812 ISSN (online):1949-4645Copyright: 2020 Pages: 3887 publication data© 2020 Published in electronic format with permission by the Society of Exploration GeophysicistsPublisher:Society of Exploration Geophysicists HistoryPublished Online: 30 Sep 2020 CITATION INFORMATION Rui Guo, Hiren Maniar, Haibin Di, Nick Moldoveanu, Aria Abubakar, and Maokun Li, (2020), "Ground roll attenuation with an unsupervised deep learning approach," SEG Technical Program Expanded Abstracts : 3164-3168. https://doi.org/10.1190/segam2020-3425792.1 Plain-Language Summary Keywordsmachine learningnoiserayleigh waveprocessingPDF DownloadLoading ...

The Feasibility of Ground-Roll Noise Suppression from 3d Seismic Data Using High-Resolution Linear Radon Transform

Ground-roll noise extraction and suppression using high-resolution linear Radon transform

Surface Waves Suppression By High-Resolution Linear Radon Transform

Ground-Roll Noise Attenuation Using a Simple and Effective Approach Based on Local Band-Limited Orthogonalization.

A Study of the Near Surface High-Frequency Compensation Technology in 3-D Seismic Exploration

Frequency Compensation Technology in Ground Penetrating Radar for Improving Resolution

Seismic Signal Denoising Using $f-x$ Variational Mode Decomposition

Ground-roll Separation From Land Seismic Records Based on Convolutional Neural Network

Application on Ultra-Shallow 3D Seismic Exploration Technology

Physics-constrained deep learning for ground roll attenuation

Separating Ground-Roll from Land Seismic Record Via Convolutional Neural Network

Ground Roll Attenuation Using Convolutional Neural Network with Dense Connected Mechanism

A 3D constrained hybrid linear-pseudo hyperbolic Radon transform

A self-supervised scheme for ground roll suppression

Optimized suppression of coherent noise from seismic data using the Karhunen-Loeve transform

Ground Rolls Removal Using Simple Bandpass Filtering Combined with Signal-and-noise Orthogonalization

Ground Roll Attenuation in Seismic Data Based on Enhanced Deep Learning Framework With Adaptive Frequency Modulation Loss

A Ground-Roll Separation Method Based on Neural Networks with Morphological Similarity Loss

Ground Roll Attenuation by Synchrosqueezed Curvelet Transform

Ground Roll Attenuation with an Unsupervised Deep Learning Approach

Ground Roll Attenuation Using Non-Stationary Matching Filtering