First-break automatic picking with fully convolutional networks and transfer learning

PreviousNext No AccessSEG Technical Program Expanded Abstracts 2019First-break automatic picking with fully convolutional networks and transfer learningAuthors: Tao XieYue ZhaoXuming JiaoWenjing SangSanyi YuanTao XieGeophysical Research Institute, China Oilfield Services LimitedSearch for more papers by this author, Yue ZhaoChina University of PetroleumSearch for more papers by this author, Xuming JiaoGeophysical Research Institute, China Oilfield Services LimitedSearch for more papers by this author, Wenjing SangChina University of PetroleumSearch for more papers by this author, and Sanyi YuanChina University of PetroleumSearch for more papers by this authorhttps://doi.org/10.1190/segam2019-3215277.1 SectionsAboutPDF/ePub ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InRedditEmail AbstractPicking first-break (FB) from seismic trace is an important step for seismic refraction and seismic reflection exploration. Conventional picking methods are mostly based on identifying the differences between seismic signal and noise in terms of amplitude, phase, or frequency. We investigate a waveform classification and FB-picking method based on fully convolutional neural networks (FCNs) and transfer learning (TL). We consider FB-picking as a binary image segmentation problem of labelling a 2D seismic image with ones on signal and zeros on noise. Through FCNs, we achieve a fast and automatic image to image waveform classification. The boundaries between noise-dominant background above FBs and signal-dominant waveforms below FBs are the location of FBs. After training FB-picking network with a training set, we transfer the optimal network parameters into another land adjacent dataset to accelerate the convergence and self-learning. Testing results demonstrate that the proposed strategy can not only achieve an efficient FB-picking result with more than 90% accuracy, but also can be flexibly extended to other seismic data.Presentation Date: Wednesday, September 18, 2019Session Start Time: 1:50 PMPresentation Start Time: 1:50 PMLocation: 301BPresentation Type: OralKeywords: traveltime, machine learning, neural networksPermalink: https://doi.org/10.1190/segam2019-3215277.1FiguresReferencesRelatedDetailsCited byConvolution neural network application for first‐break picking for land seismic data28 June 2022 | Geophysical Prospecting, Vol. 70, No. 7Automated first break picking with constrained pooling networksDavid Cova, Peigen Xie, and Phuong-Thu Trinh30 September 2020A Comparative Study of Five Networks for Reservoir Classification Based on Geophysical Logging SignalsIEEE Access, Vol. 8 SEG Technical Program Expanded Abstracts 2019ISSN (print):1052-3812 ISSN (online):1949-4645Copyright: 2019 Pages: 5407 publication data© 2019 Published in electronic format with permission by the Society of Exploration GeophysicistsPublisher:Society of Exploration Geophysicists HistoryPublished Online: 01 Aug 2019 CITATION INFORMATION Tao Xie, Yue Zhao, Xuming Jiao, Wenjing Sang, and Sanyi Yuan, (2019), "First-break automatic picking with fully convolutional networks and transfer learning," SEG Technical Program Expanded Abstracts : 4972-4976. https://doi.org/10.1190/segam2019-3215277.1 Plain-Language Summary Keywordstraveltimemachine learningneural networksPDF DownloadLoading ...